Genetic, physiologic and ecogeographic factors contributing to variation in Homo sapiens: Homo floresiensis reconsidered



    1. Human Evolution Research Center, University of California, Berkeley, CA and Department of Anatomy, University of the Pacific, Arthur A. Dugoni School of Dentistry, San Francisco, CA, USA
    Search for more papers by this author

Gary D. Richards, Human Evolution Research Center, Department of Integrative Biology, University of California, Berkeley, CA 94720, USA.
Tel.: (510) 642 7952; fax: (510) 643 8231;


A new species, Homo floresiensis, was recently named for Pleistocene hominid remains on Flores, Indonesia. Significant controversy has arisen regarding this species. To address controversial issues and refocus investigations, I examine the affinities of these remains with Homo sapiens. Clarification of problematic issues is sought through an integration of genetic and physiological data on brain ontogeny and evolution. Clarification of the taxonomic value of various ‘primitive’ traits is possible given these data. Based on this evidence and using a H. sapiens morphological template, models are developed to account for the combination of features displayed in the Flores fossils. Given this overview, I find substantial support for the hypothesis that the remains represent a variant of H. sapiens possessing a combined growth hormone–insulin-like growth factor I axis modification and mutation of the MCPH gene family. Further work will be required to determine the extent to which this variant characterized the population.


A new species, Homo floresiensis, has been recently suggested to have arisen from a Homo erectus Dubois 1892 ancestor and to have co-existed with modern humans during the Late Pleistocene (Brown et al., 2004). This announcement captured the attention and imagination of the scientific community and general public. Researchers quickly began to question the validity of the conclusions (see Balter, 2004; Diamond, 2004; Wong, 2005). The range of scenarios proposed to account for the observed morphology can be grouped such that the remains represent: (1) a new species of Homo derived from a H. erectus or earlier ancestor; (2) a pathological or anomalous individual derived from a Homo sapiens Linnaeus 1758 population; or (3) an insular dwarf evolutionarily derived from H. sapiens.

Explanations for the H. floresiensis skeletal remains must be consistent with ecological and physiological constraints on stature ontogeny in species, physiological and evolutionary constraints on brain size in species, interactions of modifications in developmental pathways (cf. pathology, anomalous development) that relate to stature and brain size in species, and correlation and meaning of features and shape characteristics of bone between modern and fossil forms.

In an effort to clarify and refocus investigation of H. floresiensis, I review the work of Brown et al. (2004) and Morwood et al. (2005) and examine the remains’ affinities with H. sapiens. Such a review is necessary in any evaluation of important issues related to stature and brain size reduction, and other taxonomic and phylogenetic issues that these discoveries raise. To investigate these aspects, I review the geographic distribution of short-statured populations, hypotheses relating to their acquisition of a reduced stature, and the genetic and physiological basis of stature ontogeny in humans. Further, I review genetic, ecological and evolutionary issues related to brain size in humans and the relationship(s) of these to anomalous and pathological restrictions of brain growth. Because the specific genetic and physiological factors related to brain growth and stature discussed herein are not fully independent, I assess the potential result(s) of their interaction. Based on these genetic, physiological, ontogenetic and ecogeographic data, I discuss the anatomy of H. floresiensis and show how its combination of ‘primitive’ and ‘derived’ features is consistent with modification of a H. sapiens morphology.

Flores remains and comparative samples

Remains reported from Liang Bua Cave, Flores, Indonesia by Brown et al. (2004) and Morwood et al. (2005) comprise an adult female partial skeleton (LB1, c. 18 kyr), an isolated lower third premolar (LB2, 37.7 to c. 74 kyr), a radius shaft (LB3, c. 74–95 kyr), an incomplete subadult radius and tibia (LB4, c. 12 kyr), a mandible and associated fragmentary post-cranial remains (LB6, c. 15 kyr), a mostly complete tibia (LB8, c. 18 kyr), and other miscellaneous remains (LB5, LB7 and LB9). All Pleistocene remains described by Brown et al. (2004) and Morwood et al. (2005) are now referred to H. floresiensis. Morwood et al. (2005) note the existence of newly recovered remains that they attribute to H. sapiens. These H. sapiens remains are said to derive from Holocene deposits directly above the latest occurrence of H. floresiensis (=LB4).

Brown et al. (2004) and Morwood et al. (2005) compare dimensions of the LB1–2 and LB6 craniodental remains to a large, geographically and temporally diverse sample of H. sapiens. These metric data derive from numerous sources and are combined by Brown et al. (2004) into a single H. sapiens series. The metric sample employed by these authors for analysis of the post-cranial skeleton is unspecified, but reference to Brown (1998–2004) suggests that it is sizable but less geographically diverse than the skull sample. Note that Morwood et al. (2005) utilized a small sample of limb dimensions for African pygmies (n = 10) and Andaman Islanders (n = ?) for comparative purposes. Excepting the latter few comparisons, conflation of these data sets eliminates the ability to correlate observed features of LB1–LB9 with those of geographically delineated H. sapiens populations and effectively moves the analysis to the species level. Both Brown et al. (2004) and Morwood et al. (2005) also provide metric and nonmetric comparisons with remains attributed to H. erectus, early Homo (cf. H. habilisLeakey et al., 1964), Australopithecus, great apes and monkeys.

Relations of Flores skeletal features to Homo sapiens

Based on metric comparisons, the LB1 individual is described as presenting a low cranial capacity (380–417 cm3), short stature (c. 106.0 cm: Brown et al., 2004), and both a wide pelvis and long arms in relation to its legs, when compared with H. sapiens (Aiello, in Balter, 2004; Brown et al., 2004; Falk et al., 2005a; Morwood et al., 2005). These observations are important as they suggest a resemblance to some early hominids (A. afarensisJohanson et al., 1978 and H. habilis) and clearly played a significant role in the interpretation and assignment of the LB1–2 (Aiello, in Balter, 2004; Brown et al., 2004) and LB4–9 remains (Morwood et al., 2005).

Comparison of the estimated brain size and stature of the LB1 individual to those for H. sapiens shows them to be outside the normal range of variation. They are not, however, outside the known range for viable individuals within the species (350 to c. 2000 cc; 57–272 cm, respectively: contra Brown et al., 2004). Further, research on surface area to body mass ratios makes it clear that stature is highly variable in modern humans, and reference to modern pygmy populations shows stature to be subject to significant and rapid change (Hiernaux et al., 1975; Hiernaux, 1977; Bailey & DeVore, 1989; Ruff, 1994). Given this fact, it follows that Beals et al. (1984) found only a weak correlation between stature and cranial capacity in H. sapiens, although Holloway (1980) found a within-population correlation in a Danish sample. Further, based on their global assessment of cranial capacity, Beals et al. (1984) conclude that brain size should only be accorded the same weight as any other variable in taxonomic assessments of Pleistocene hominids.

Because of the broad and synthetic focus of the current work, only a subset of the cranial and post-cranial features ascribed to H. floresiensis are specifically discussed below, whereas a summary is provided for a compiled feature list. This subset of features is considered sufficient to demonstrate the need for population-level comparisons with H. sapiens.

Brown & Morwood (2004) have recently argued that the LB1 mandible possesses a combination of nine features that eliminates any possibility it could derive from a modern human. At first glance this feature set would seem to represent ‘…component parts of a total morphological pattern of associated parts…’ (LeGros Clark, 1964, p. 16–17) that would characterize a species-level difference. However, seven of the nine features are normal variants in modern humans, generally, and Indonesian, Malaysian and African groups, specifically (Gregory, 1920; Keiter, 1933; Genet-Varcin, 1951; Sauter & Adé, 1953; Verhoeven, 1958; Jacob, 1967; Marquer, 1972; Birdsell, 1993). The two remaining features, absence of a chin and presence of well-developed superior and inferior transverse tori, not normal H. sapiens variants, are of interest. A significantly reduced chin (vertically aligned corpus and alveolar units) and developed transverse tori are, however, features found in some African and Indonesian pygmies (Jacob, 1967; Marquer, 1972) and Australomelanesians (Keiter, 1933; Birdsell, 1993) respectively. The significance of the vertically aligned chin region is discussed below.

The LB1–LB9 post-cranial remains are also considered by Brown et al. (2004) and Morwood et al. (2005) to be unique relative to H. sapiens. Three features are of particular interest: (1) humeral head torsion angle; (2) humeral, femoral and tibial shaft circumference relative to length and (3) pelvic shape.

Morwood et al. (2005, p. 1016) determined a humeral head torsion angle of c. 110° for LB1 and noted that this is ‘… the norm for Hylobates and quadrupedal primates such as Macaca, but is significantly less than in large-bodied apes, modern humans (141°–178°) and other known hominids, including Australopithecus’. The important points related to this assertion are that: (1) modern pygmies are known with similar angles and have low torsion angles for H. sapiens, in general; (2) the published range for this angle is 111°–187° in modern humans; (3) such low torsion angles are not the norm for hylobatids, generally; (4) high humeral head torsion angles are potentially related to the evolution of increased manipulative abilities in humans and (5) the humeral head of LB1 is damaged and this may have impacted measurement of the angle (Broca, 1881; Le Damany, 1903; Matiegka & Maly, 1938; Evans & Krahl, 1945; Krahl & Evans, 1945; Genet-Varcin, 1951; Sauter & Könz, 1954–1955; Jacob, 1967; Marquer, 1972; Larson, 1988; Eldelson, 1999, 2000).

Unusually large values for the humeral, femoral and tibial circumferences in LB1, given the short length of the shaft (diaphysis), have also been reported by Morwood et al. (2005). They observe that these values place the humerus midway between their Pan paniscus and H. sapiens samples, the femur within the P. paniscus range, and the tibia within the Pongo and Pan ranges. These data are used by Morwood et al. (2005) to suggest that estimations of musculature and body mass in H. floresiensis would be more accurate if based on chimpanzee rather than H. sapiens models. Whereas such an assessment is of interest, of greater concern are: (1) that a determination of the animal's locomotor pattern is required prior to undertaking body size determinations (Damuth & MacFadden, 1990; Ruff, 2003a) and (2) the implications of such circumference-to-length ratios for locomotor behaviour.

It is well known that limb bone growth maintains an ‘elastic similarity’ which predicts that bone diameter should scale to bone length, except during development. It is also known that diaphyseal cross-sectional areas reflect both physical activity (resistance to bending and torsional loads) and body size (Sumner & Andriacchi, 1996; Gasser et al., 2001; Ruff, 2002, 2003b, 2005; Janz et al., 2004; Rauch, 2005). Further, because the diaphyses are responding to both mechanical loading and systemic physiological requirements, the thickness of the shaft's cortical bone and medullary canal size are critical factors in understanding both locomotor activity and body size (Ruff, 2002). In any assessment of these elements, it is important to recall that the trunk–forelimb and hindlimb regions may be partially dissociable growth units and would, therefore, be expected to show differences if the forelimbs and hindlimbs are used differently during locomotion and support of the body mass (Shea, 1983; Gasser et al., 2001; Ruff, 2002).

In H. floresiensis, the humeral circumference-to-length values appear to be more significantly increased relative to H. sapiens, whereas the femoral and tibial values appear less dramatically increased. Given these data, two possibilities present themselves: (1) that the shaft diameters are responding to increased locomotor loading, especially those of the upper limb; or (2) that the quality of the bone comprising the shaft's cortex is of inferior quality, necessitating a greater diameter in order to accommodate normal loads.

Brown et al. (2004) state that H. floresiensis lacks the reduced medullary canals of earlier hominids and, by extension, a thickened shaft cortex. If bone comprising the diaphyses of H. floresiensis is ‘normal’ and granting that external shaft dimensions are not as useful as section moduli (Ruff, 2002), it would appear that the Flores individuals engaged in either or both a significant amount of suspensory (i.e. climbing) activity or semi-quadrupedal locomotion. Clearly such a potential difference in locomotor behaviour has important implications for body size estimates, postures and other behaviours attributed to this species (i.e. hunting, tool making). Because modern short-statured populations possess physiological characteristics which impact factors related to bone growth, it is important to understand the relationship between stature reduction, locomotion, and body size in these groups prior to undertaking an assessment of H. floresiensis remains.

In considering pelvic shape, Morwood et al. (2005) have indicated that the ilium of LB1 is anterolaterally flared and consistent with an australopithecine-shaped thoracic region, as opposed to the shape found in H. sapiens. Whereas quantification or comparisons of these suggested similarities are unavailable, it is important to recall that pelvic and femoral anatomy are closely correlated (Ruff, 2005) and also reflect locomotor adaptations. Homo  floresiensis shares the majority of its femoral anatomy with H. sapiens and not earlier hominids, inclusive of Australopithecus. If a significant lateral flare is characteristic of H. floresiensis, it may be related to the significant limb size reduction and, potentially, associated physiological changes related to stature reduction in H. sapiens.

By compiling a list of 90 skeletal features from discussions in Brown et al. (2004) and Morwood et al. (2005), this author found that 77.8% (70/90) either occur in H. sapiens generally or are known variants with more variable frequencies of occurrence. Also, many of these features are reminiscent of those found in various modern and prehistoric Australo-Melanesian populations, as also observed by Henneberg & Thorne (2004). In some features, H. floresiensis is at or slightly below the lowest end of the H. sapiens range of variation, as would be expected given its small brain and body size. Of the 20 features that are not normally found in H. sapiens, the majority are related to brain size reduction and modifications of ontogenetic pathways, as discussed below.

Modern and fossil pygmies

Given that H. floresiensis is characterized by an extremely short stature, suggested to have resulted from insular dwarfism (Brown et al., 2004), comparison to dwarfed H. sapiens populations is appropriate. Whereas insular dwarfing implies an island environment, I follow Cavalli-Sforza (1986a) in viewing mainland tropical forests as ‘ecological islands’. The validity of this view is supported by the fact that no significant difference has been demonstrated in epigenetic mechanisms, phenotypic results, or likely selective forces between short-statured populations from island and mainland tropical forest habitats (Cavalli-Sforza, 1986a; Merimee & Rimoin, 1986; Merimee et al., 1987; Schwartz et al., 1987; Shea et al., 1992; Clavano-Harding et al., 1999; Dávila et al., 2002). It is for this reason that the term ‘pygmy’ is simply another term for physiological dwarfing.

Whereas the basic mechanism(s) that results in short statured human populations is similar, it does not follow that all reduced stature populations are the same. Selective pressures universal to H. sapiens tend to result in similar cranial capacities and, consequently, pelvic sizes in these populations. Analysis of African short-statured populations indicates, however, that each differs in unique ways, particularly by variation in limb segments, from adjacent human populations (Hiernaux, 1977; Smith & Buschang, 2005). These differences in specific features of the cranial and post-cranial skeleton are thought to be responses to the equatorial climate (von Eickstedt, 1931; von der Brock, 1939; Hiernaux, 1977).

Geographically dispersed pygmy populations differ slightly but do possess specific and universal features of the post-cranial skeleton. Based on studies of climatic adaptations in hominids, Ruff (1994) found body breadth to be relatively stable and genetically canalized, compared with stature or body weight. His observations are consistent with results provided by Vincent et al. (1962) and Cavalli-Sforza (1986b), both of whom found the pelvis of African pygmies to be slightly small relative to other Africans but to be less reduced in width when compared with the decrease in stature. These data predict that the LB1 pelvis would be ‘wide’, if the population from which it derived underwent a reduction in stature similar to or greater than that in modern human pygmies.

Marquer (1972) found the intermembral index in Western and Eastern African Pygmies to show a pronounced development of the upper limb relative to the lower limb. A similar relationship in the intermembral index has also been observed by Trinkaus (1981) and Shea & Pagezy (1988). Both Shea & Pagezy (1988) and Shea & Bailey (1996) indicate that body proportions in pygmies are what would be expected given a generalized allometric truncation or ontogenetic scaling of the growth cycle. Recent description of the upper limbs of LB1 allows confirmation of a pygmy pattern in upper-to-lower arm proportions and arm-to-leg proportions (i.e. long arms and short legs) in combination with upper and lower limbs that are short relative to modern pygmies (Morwood et al., 2005).

Modern human pygmy and pygmoid1 populations are widely distributed globally and their short stature appears to represent one aspect of a complex ecogeographic adaptation to rain forest or island environments (Coon, 1965; Brues, 1977; Cavalli-Sforza, 1986a; Ruff, 1994; Shea & Bailey, 1996; Katzmarzyk & Leonard, 1998). Note that alternative or contributory factors related to the acquisition of short stature also include limited food resources and genetic drift (Diamond, 1992; Shea & Bailey, 1996). Regardless of the factor(s) involved, all short-statured populations share an apparent disruption of the growth hormone–insulin-like growth factor I (GH–IGF-I) axis. While noting both local and regional variation in pygmy populations, short-statured populations in Africa and Southeast Asia have a combined range for stature of c. 126.2–172.1 cm for males and c. 116.5–163.0 cm for females (Table 1). The Efe and Mbuti pygmies, who inhabit the Ituri forest of Zaire, are considered to have the shortest statures for any living human group, although a normal Aka female possessed a stature of only 116.5 cm (Flower, 1889).

Table 1.   Stature estimates (in cm) for African and Southeast Asian modern human pygmies.
  1. *Data derived from the following sources: (a) Cavalli-Sforza (1986b); (b) Marquer (1972); (c) Vincent et al. (1962); (d) Bailey (1991); (e) Flower (1889); (f) Pennetti et al. (1986); (g) Dávila et al. (2002); (h) Clavano-Harding et al. (1999); (i) Headland (1989); (j) Newton (1920); (k) Schwartz et al. (1987).

  2. Stature measured directly from living individuals.

  3. Stature range calculated from the author's sample mean and represents ±3 SDs.

  4. §Stature is estimated from skeletal remains.

West African Pygmies*
Bagandu, Central African Republica†
 Mean152.7 ± 1.35145.0 ± 1.18
Ba-Bongo, Gabon; Ba-Binga, Central African Republic
Ba-Bongo, Gabon; Ba-Binga, Central African Republic b†
Central African Republic b†
 Mean152.9 ± 6.4144.3 ± 5.9
East African Pygmies*
Bayenga pygmies, Democratic Republic of the Congoc†
Ngayu, Democratic Republic of the Congoc†
Epulu Region, Ituri forest, Democratic Republic of the Congo
 Mean144.4 ± 0.54136.0 ± 0.72
Efe, Ituri Forest, Democratic Republic of the Congod†
Epulu Region, Democratic Republic of the Congoa†
 Mean144.2 ± 6.0137.4 ± 4.37
Efe, Basa and Aka
Efe and Basua f†
Aeta Negritosg
Ati Negritosh
 Mean149.0 ± 7.0144.0 ± 3.0
Casiguran Agta Negritosi
Mamanwas Negritosg
New Guinea*
Mountain Ok, Papuak†
 Mean152.0 ± 6.0146.0 ± 5.7

Based on their extremely short stature and the extent of gene frequency divergence from the average African, Mbuti pygmies are considered the most pure or least admixed (Hiernaux, 1977; Cavalli-Sforza, 1986a). Of interest is that Hiernaux (1977) suggests that the extreme stature reduction of the Mbutis must result from a great time depth in a rain forest environment, as other groups in similar environments do not share it. Alternatively, because pygmy populations vary slightly in the way that they achieve a short stature, one must consider the possibility that the stature of Mbuti pygmies is due to either or both a population-specific upregulation (increased expression) of factors involved in the physiology of body size or a difference in the onset of these factors.

For island pygmies, the Ati Negritos of the Philippines possess the shortest statures (Table 1). The Ati live in rain forest habitats and are, therefore, similar to both mainland pygmies and other Southeast Asian island dwellers. Given the known effect on body size associated with isolation in island environments it is possible that rain-forest-adapted populations of pygmies shorter than the Efe, Mbuti, or Aka existed in island environments but are undocumented. Currently, no data exists for island groups as to the degree of stature reduction related strictly to an isolated island context and that related to life in mainland rain forests. Clarification of this problem will be complicated as research on both African and Southeast Asian pygmies shows extensive and somewhat obligate interconnections to exist with surrounding nonforest-adapted groups (Cavalli-Sforza, 1986a; Wijsman, 1986; Headland, 1989; Headland & Reid, 1989) and these interactions are now known to be moderating the genetic effects of ‘isolation’ in modern pygmy populations (Katzmarzyk & Leonard, 1998).

In prehistoric contexts, short-statured individuals are known from numerous localities in Southeast Asia, generally, and Liang Toge Cave, Flores, Indonesia, specifically (Verhoeven, 1958; Jacob, 1967). The latter is a nearly complete adult female skeleton (Verhoeven, 1958). This specimen is dated to 3.55 ± 0.525 kyr BP (Jacob, 1967). Further, it would be useful to have information on any hominid remains recovered from Liang Bua Cave by T. Verhoeven in 1965 and R. P. Soejono in 1978–1989 (Morwood et al., 2004) and the Holocene remains attributed to H. sapiens by Morwood et al. (2005). Currently, any potential data on these remains appears to be unavailable in the published literature.

Brown et al. (2004) note the existence of modern pygmy groups but do not make direct comparisons. They make no mention of pygmoid or proto-pygmoid remains from prehistoric contexts. Morwood et al. (2005) do provide a limited set of post-cranial comparisons with African and Andaman Island short-statured populations. Brown et al. (2004) provide two reasons for excluding pygmies from their comparisons: (1) known pygmies possess brain sizes and craniofacial proportions within the range of adjacent larger-bodied populations (Brown et al., 2004) and (2) the ‘…combination of small stature and brain size in LB1 is not consistent…’ with IGF-I ‘…related postnatal growth retardation…’ (Brown et al., 2004, p. 1060). Morwood et al. (2005, p. 1016) expand this theme by suggesting that the body proportions of LB1 are not abnormal as ‘…growth-hormone-related dwarfism and microcephaly in modern humans result in normal limb and pelvic proportions’.

Although modern short-statured populations have brain sizes which are consistent with other H. sapiens found in similar ecogeographic regions (Beals et al., 1984), they are known to be reduced relative to the H. sapiens average (Genet-Varcin, 1951; Vincent et al., 1962). Further, the use of craniofacial proportions in this context is misleading as it is based on purely subjective criteria. Modern short-statured populations possess physiological differences from other human groups and the ontogenetic effects of these differences extend to the craniofacial skeleton. Note that Daughaday & Harvey (1995) discuss ontogenetic modifications of the craniofacial region related to disruption of the GH–IGF-I axis in nonpygmy individuals. Details of the physiological basis for these differences will be discussed below. Further, as noted above, modern pygmy populations possess minimum stature estimates for females that approximate (within 10–20 cm) that of the LB1 and LB8 individuals. As it is known that brain size is only weakly correlated with stature (Beals et al., 1984), rejection of direct comparisons with pygmy populations based on brain size differences is inappropriate.

Brown et al. (2004) and Morwood et al. (2005) fail to discuss or examine in detail known short-statured populations. However, the inference that a dramatic reduction in human growth would be required to result in the short stature of the LB1 and LB8 individuals is not well warranted. Only some of the rarest skeletal dysplasias result in isolated high magnitude growth restrictions in humans. Brown et al. (2004) observe this fact and conclude that neither pituitary dwarfism nor ‘primordial microcephalic dwarfism’ (PMD) are consistent with the morphology of the LB1 individual [note that the term ‘PMD’ has no validity in current clinical contexts and that the condition Brown et al. (2004) are referring to is microcephalic osteodysplastic primordial dwarfism; MOPD: Online Mendelian Inheritance in Man; OMIM 210710–210730: Majewski & Goecke, 1982].

Individuals with MOPD share some features with the LB1 individual and vary from relatively normal (Halder et al., 1998) to severely malformed (Majewski & Goecke, 1998; Kraft et al., 2000; Fukuzawa et al., 2002; Klinge et al., 2002; Maclean et al., 2002; Nishimura et al., 2003). This is also the case for pituitary dwarfism. It is not necessary, however, to postulate the presence of such high magnitude malformation syndromes. When the LB1 and LB8 stature estimates are compared with stature in small-bodied modern and prehistoric humans, the differences are not so dramatic. Such a difference in stature could reflect either or both: (1) a continuation of stature reduction in H. sapiens pygmies within rain-forest habitats, resulting in evolutionary diminution in stature; or (2) an anomaly in one or more individuals that results from variant developmental or pathological processes. These two possibly overlapping causal mechanisms are worthy of investigation before entertaining alternatives requiring deeper phylogenetic divisions.

Normal and abnormal brain development

When direct comparisons of the LB1–LB9 remains with modern and prehistoric short-statured populations are conducted, certain features of the skull of LB1 emerge as unique. These features include very small brain size, cranial shape characteristics, lack of a chin, and, possibly, aspects of the cranial base and dentition (Brown et al., 2004; Lahr & Foley, 2004). The describers indicate that neither the physiological characteristics of modern pygmies nor the reduced brain and body size of pituitary dwarfism and MOPD are consistent with the morphology of LB1–LB9.

Modern short-statured human populations share an apparent disruption of the GH–IGF-I axis that does not appear to affect their brain development, although data to support such a conclusion are not available. Individuals with pituitary dwarfism and MOPD, while presenting clinical features of short stature and variably reduced brain sizes, are rare and present morphologically inconsistent patterns. Reference to either of these groups will not, therefore, solve the problem of the small brain size of H. floresiensis. This fact spawned suggestions that the LB1 skeleton could represent an abnormal or ‘diseased’ (i.e. pathological) individual with short stature and a small head (microcephaly). Whereas Brown et al. (2004) rejected the latter possibility, it was revived by both Henneberg & Thorne (2004) and Jacob (2004).

Henneberg & Thorne (2004) compare the LB1 skull to an archaeologically derived ‘microcephalic’ from prehistoric Crete (Poulianos, 1975), concluding that the craniofacial relationships and receding chin of LB1 are consistent with a diagnosis of secondary microcephaly. Antón (in Wong, 2005, p. 57) countered that ‘… the facial morphology is completely different in microcephalic humans [compared to LB1], and their body size is normal, not small’. Paradoxically, she concluded that the LB1 individual ‘could be a H. erectus individual with a pathological growth condition stemming from microcephaly or nutritional deprivation’. Falk et al. (2005a,b) attempt to counter suggestions that the LB1 individual is a microcephalic (sensu pathological development) by comparing virtual endocranial casts of LB1 and a modern human claimed to have primary microcephaly or ‘microcephaly vera’ (Falk et al., 2005a). These authors reject the possibility that LB1 was afflicted with either primary or secondary microcephaly and conclude that the shape of the brain is most similar to H. erectus.

Prior to interpreting the results of these studies of gross brain anatomy and reviewing aspects of brain development, it is important to recognize that: (1) the term ‘microcephaly’ refers simply to the condition of having a small head relative to age- and sex-corrected means (Mochida & Walsh, 2001; Woods et al., 2005) and (2) criteria applicable in clinical settings cannot be fully applied to skeletal remains (Richards, 1985). Factors resulting in microcephaly may be primary (occurring early in development) or secondary (occurring later in development). In both instances the condition is heterogeneous, presenting with widely varying morphological and behavioural features. This heterogeneity is reflected in the c. 300 listings of genetically induced conditions that include primary microcephaly in the diagnosis (

Efforts to delineate specific conditions in clinical settings and, by extension, in paleopathological research have been hampered by the wide range of factors that result in a vaguely similar skull phenotype (i.e. small: Komai et al., 1955; Richards, 1985; Mochida & Walsh, 2001). Much of this confusion results from ‘diagnoses’ made by the simple application of deviations from the normal occipitofrontal circumference (OFC) wherein ≤3.0 SDs equates to a ‘diagnosis’ of primary and ≤2.0 SDs equates with secondary microcephaly. In clinical settings, the OFC is an indicator of potential problems; it is not a diagnosis. In the case of skeletal remains, Richards (1985) demonstrates that a complete analysis of the skull, brain endocast, and post-cranial skeleton are necessary when considering a ‘microcephalic’ individual in order to limit the number of potential causative factors and provide a listing of alternative diagnoses (differential diagnosis).

Unfortunately, in their analysis of the LB1 individual Falk et al. (2005a) acquired a skull with a reduced brain size and concluded that it represents an individual with primary microcephaly or ‘microcephaly vera’. Based on this single, abnormally configured brain Falk et al. (2005a) reject the hypothesis that the LB1 adult was a microcephalic (sensu pathology). They also assign specific morphological criteria (derived from a website overview) to this condition. It is important to note that Falk et al. (2005a): (1) employed an endocranial cast made from a poor-quality replica of the skull of Jakob (Jacques) Mœgle, a 10-year-old individual; (2) failed to state that, with a cranial capacity of 272 cc, this individual possessed one of the smallest known human brain sizes and (3) failed to reference the individual's medical history, inclusive of a full autopsy report (Vogt, 1866). Referencing aspects of this individual's medical history within OMIM returns a limited set of potential genetic syndromes. Individuals afflicted with these syndromes present microcephaly as one feature of a much broader set of developmental defects; in other words, this individual was not simply a microcephalic [sensu stricto (s.s)]. Further, reference to the primary literature reveals that the term ‘microcephaly vera’ has been superseded and that ‘primary microcephaly’ simply refers to the time in development when the ‘microcephaly phenotype’ is acquired.

Currently, autosomal recessive primary microcephaly (MCPH1 to 6) is the condition proposed to replace the outdated, nonspecific, and overly misused synonyms ‘microcephaly vera’, ‘true microcephaly’ and ‘primary microcephaly’ (when used as a specific diagnosis: Komai et al., 1955; Richards, 1985; Roberts et al., 2002; Woods et al., 2005). Giacomini's (1890) original intent in proposing the term ‘microcephaly vera’ is retained, but the condition is differentiated from other similar conditions, such as Seckel syndrome, by: (1) a direct link to specific genetic loci and genes and (2) diagnostic criteria that include only a small brain size (=cerebral cortex; brain weight c. 430 g, but ≥4 SD below age and sex means: McCreary et al., 1996) and bony neurocranium and a lack of syndromic features or significant neurological defects other than mild to moderate mental retardation. Of interest is the observation of Roberts et al. (2002) that even the classic sign of a sloping forehead is not applicable to all MCPH cases. Recently, Woods et al. (2005) have expanded this definition to include findings that some MCPH1 individuals present with reduced stature, although they note that the impact on height is always less than the impact on head circumference.

Because the morphology of the LB1 skull is unlikely to be the product of one of the secondary causes of microcephaly (contra Henneberg & Thorne, 2004), it is necessary to focus on primary causes. Given that discussions of this new species derive, in part, from observations of the surface anatomy of brain endocasts (Falk et al., 2005a,b; Weber et al., 2005; Holloway et al., 2006), it is important to note that primary microcephaly can be split into two subcategories, MCPH and microcephaly with simplified gyral pattern (MCsg: Mochida & Walsh, 2001; Sztriha et al., 2004). Whereas individuals with MCsg have gyral and other architectural abnormalities, these are subtle and differentiate this condition from more extreme forms such as type I lissencephaly (Mochida & Walsh, 2001). The latter authors note that such conditions are diverse, genetically heterogeneous and potentially overlap conditions such as MCPH. Mochida & Walsh (2001) confirm this observation by noting that, whereas the gyral pattern in MCPH is relatively normal, it does exhibit some difference due to the smallness of the brain. Whereas the gyral-sulcal pattern may show some slight differences, an abnormality of cortical laminar formation has not been detected at the microscopic level in MCPH. In some cases, however, depletion of ‘later born’ neurons in cortical layers II and III, as well as early depletion of precursor cells in the germinative zone near the ventricles, has been observed (Mochida & Walsh, 2001). Recently, Falk et al. (2005b) have suggested that, contrary to Weber et al. (2005), normal gyral patterns are believed to be typical of ‘true’ microcephalics, whereas simple gyrification typifies some kinds of secondary microcephaly. Given the previous discussion, such an assertion cannot be supported.

The MCPH gene family and brain size

Brain size, or cranial capacity, is a central focus for discussion of the taxonomic validity of H. floresiensis. Because of rapid developments in identifying genes related to brain development a significant body of new genetic data is now available. The most relevant aspects of these new data are presented below whereas a detailed overview is provided in electronic Appendix S1.

Six recessive MCPH loci have already been mapped (MCPH1–MCPH6: Mochida & Walsh, 2001; Leal et al., 2003). Individuals with these gene abnormalities present with a generally similar phenotype characterized by a significantly reduced cerebral cortex and mild to moderate mental retardation but a lack of other significant clinical symptoms (Mochida & Walsh, 2001; Roberts et al., 2002; Bond et al., 2003; Leal et al., 2003). In MCPH1 (OMIM 251200), there is a truncating mutation of the microcephalin gene. Microcephalin is expressed in the developing forebrain, especially in the walls of the lateral ventricles. MCPH5 is caused by mutation of the human abnormal, spindle-like, microcephaly-associated gene (ASPM: OMIM 650481). ASPM shows its greatest expression in that portion of the ventricular zone containing progenitor cells for cerebral cortical pyramidal neurons (Bond et al., 2002). Neural progenitor cells arise in the ventricular region through both symmetric and asymmetric mitosis to produce neurons that migrate to form the cerebral cortex (Rakic, 1995).

In MCPH, MRI brain scans show that the whole brain is reduced in size and that it is the symmetrical reduction of the cerebral cortex, due to a decrease in the number of neurons, that results in the small size (Bond et al., 2002, 2003; Roberts et al., 2002). This reduction in neurons can result from: (1) decreased proliferation of neuronal progenitors during symmetric mitosis; (2) decreased production of mature neurons by each neuronal progenitor during asymmetric mitosis; or (3) excessive apoptosis of neuronal progenitors or mature neurons (Mochida & Walsh, 2001).

Regression or degeneration of established neural tissue can result in small brain size, but MCPH appears to arise through variation in the normal developmental process (Roberts et al., 2002). The degree to which the cerebral cortex is reduced in MCPH varies. Bond et al. (2003) report that, in 23 consanguineous families expressing MCPH5, 19 families presented with reduced OFCs of 5.0–11.0 SDs and mild to moderate mental retardation, whereas four families presented with reductions of 10.0–11.0 SDs in association with severe mental retardation. Roberts et al. (2002) found reduced OFCs of 4.0–14.0 SDs with individuals in families varying by only ±1.0 SD. Of interest is that these authors were able to compare the OFCs of fifteen subadults at birth and at 5.0–10.0 years of age. Their finding of no increase in the OFC in these individuals implies that the condition is fully manifested at birth (Roberts et al., 2002). Whereas one might expect such reduced brain sizes to be associated with other somatic and behavioural problems, Roberts et al. (2002) observed that all individuals were in good health and displayed normal early motor milestones (although some had delayed speech). None of these individuals had severe or profound mental retardation, and most were physically able and trainable in life skills (Roberts et al., 2002). These observations are consistent with those for the majority of individuals in the Bond et al. (2003) study. Jackson et al. (2002) note that the ways in which mutations of microcephalin and ASPM cause a reduction in brain size in the apparent absence of other major phenotypic effects will require further investigation.

In considering the relationship of brain size and stature it is clear that microcephaly, generally, and MCPH, specifically, can occur in individuals of normal stature. However, both Komai et al. (1955) and Okuhara et al. (2000) suggested that microcephaly [sensu latu (s.l.)] could be linked to a reduction in stature. Vargas et al. (2001) furthered this observation by noting that the co-occurrence of primary microcephaly (s.l.) and normal stature is rare. The recent re-definition of MCPH (Woods et al., 2005) confirms these earlier observations of stature reduction in individuals with microcephaly (s.l. and s.s.: contra Antón, in Wong, 2005).

It is generally recognized that the incidence of MCPH offspring is about twice as high among consanguineous than nonconsanguineous couples (Rittler et al., 2001). Recognition of this correlation has allowed delineation of the MCPH gene family (Mochida & Walsh, 2001; Bond et al., 2002; Jackson et al., 2002; Roberts et al., 2002; Bond et al., 2003, 2005). Results of these works provide an incidence rate for MCPH that ranges from 1/30 000 to 1/250 000 live births in western populations (Kumar et al., 2002). In current contexts, the work of Kloepfer et al. (1964) on three consanguineous groups from rural Louisiana is of interest. An incidence rate of primary microcephaly (s.l.) in these populations was calculated to be 1 : 1000 live births. In a similar study, Qazi & Reed (1975) determined that 79 pregnancies in 13 consanguineous Canadian families resulted in 61 live births of which 28 (45.9%) presented with primary microcephaly (s.l.). Further, the works of Böök et al. (1953), Kloepfer et al. (1964) and Qazi & Reed (1975) indicate that heterozygotes have reduced intelligence. The latter authors indicate that subnormal intelligence was present in c. 50% of the heterozygotes. Alternatively, in carriers of the mutated MCPH genes, neither Roberts et al. (2002: MCPH) nor Bond et al. (2003: MCPH5) observed diminished intelligence as a heterozygote effect.

Growth hormone–insulin-like growth factor I axis

It is clear that significantly reduced brain sizes can be produced by a group of reasonably common genetic changes that affect neuronal precursor cells and result in reduced cerebral cortex size. Furthermore, a direct correlation between the degree of brain size reduction and the degree to which mental capacity is impaired in these cases does not appear to exist. These data confirm that human brain sizes in the range suggested for H. floresiensis are not necessarily incompatible with the acquisition of basic survival skills. They also indicate that such malformations are highly likely to arise in small, isolated populations such as those inhabiting remote islands.

There is also reasonable evidence for a generalized delay in somatic development in MCPH individuals. The mechanism resulting in this delay is currently unknown but may be related to recent indications that MCPH genes encode multiple gene products (Andreadis, 2005; Kouprina et al., 2005: electronic Appendix S1). Because small, isolated populations in island contexts are likely to practice consanguinity, be more susceptible to genetic drift, and be the target of selection for small body size, it is of interest to examine this body size adaptation, the role of the GH–IGF-I axis in growth, and the potential for an interaction between the GH–IGF-I axis and MCPH related protein expression.

Recent years have seen a significant increase in research on the both sides of the GH–IGF-I axis. These works provide a broad and intricate body of data on the role of these axis components in ontogenetic processes. Because of: (1) the amount and complexity of available data; (2) the integral roles played during ontogeny by the components of the axis, their carriers, and their receptors; (3) the overlap in effect on neurogenic and somatic growth between MCPH and microcephaly (s.l.) and (4) the implied and disputed role(s) of the GH–IGF-I axis in pygmy stature, an overview of only the most relevant aspects of findings involving the axis are provided below. A significantly more detailed discussion on relevant aspects of the GH–IGF-I axis is provided in electronic Appendix S2. Individual changes or a combination of changes in these axis components might result in the ‘unusual’ combinations of functional cranial components in endemic dwarfing situations, as noted by Brown et al. (2004).

Growth hormone

The GH–IGF-I axis is of interest, in part, because GH is one of three main endocrine systems playing key roles in postnatal growth, the other two being insulin and thyroid hormones (Foucher et al., 2002; Sanders & Harvey, 2004). Growth hormone is released into the circulatory system from the pituitary gland and stimulates receptors in nearly all body tissues, organs and cells (Harvey & Hull, 1995; Bennett & Robinson, 2000). Whereas GH levels are high during fetal life, pituitary-derived GH has not been considered to be required for fetal development. Recent evidence demonstrates, however, that the neural-derived (short-loop) production of GH during prenatal development has a direct action on neural development prior to the production of pituitary somatotrophs (Sanders & Harvey, 2004). This neural-derived GH has been shown to have a number of neurotrophic actions that include stimulating neuronal and glial cell proliferation and increasing myelination (Bennett & Robinson, 2000). It is now well established that GH affects both brain and cranial size in young animals (Harvey, 1995; Bennett & Robinson, 2000).

Binding of GH to receptors (GHR) can occur directly, or it may be mediated by binding proteins (GHBP: Scans & Campbell, 1995; Bennett & Robinson, 2000: electronic Appendix S2). A progressive rise in GHBP levels occurs during infancy, the prepubertal period and during adolescence. The largest increase occurs during the first few postnatal years. The ontogenetic increase in GHBP concentrations, as indicated by in vitro studies, may provide a mechanism by which growth can be attenuated (Harvey & Hull, 1995), as there is a positive correlation throughout ontogeny between overall body mass and height and serum levels of GHBP.

Insulin-like growth factor I

The other component of the GH–IGF-I axis comprises the insulin-like growth factors. These growth factors are a peptide family with diverse metabolic roles that include mediating many of the anabolic and mitogenic actions of GH (DiMeglio et al., 2000: electronic Appendix S2).

The critical role played by IGF-I during prenatal ontogeny is well established and currently considered to be more important than that of GH (Bailey, 1990, 1991; Guevara-Aguirre, 1996; Woods et al., 1996; DiMeglio et al., 2000; Bunn et al., 2005; Walenkamp et al., 2005). The effects of GH in regulating IGF-I expression during postnatal ontogeny are well established (Merimee et al., 1987; Guevara-Aguirre, 1996; Jain et al., 1998; Dentremont et al., 1999; D'Ercole et al., 2002; Foucher et al., 2002; Walenkamp et al., 2005). Based on recent evidence it appears, however, that the developmental expression and functions of IGFs may be either GH-dependent or GH-independent (Sanders & Harvey, 2004). In general, the mode of action of IGF-I appears to be as a developmental differentiation and proliferation factor for the morphogenesis of various tissues, including muscle, skin, bone, and the nervous system (Popken et al., 2004; Sanders & Harvey, 2004). Given this widespread effect, it follows that Reece et al. (1994) and Verhaeghe et al. (1993) found a direct correlation between neonatal weight and cord serum IGF-I levels. More specifically, there is now good evidence that IGF-I can both promote proliferation of neural cells in the embryonic central nervous system in vivo and inhibit their apoptosis during postnatal life (Guevara-Aguirre, 1996; Popken et al., 2004; Sanders & Harvey, 2004).

Disruption of the GH–IGF-I axis

Most data on the role of the GH–IGF-I axis in growth derives from individuals with mutations or dysfunction of the coding sequence. Because of the complexity of the GH–IGF-I axis, only a small subset of the better-known defects in this system can be presented in the following overview. Conditions presented should not be considered as potential diagnoses for the H. floresiensis individuals but as indicators of the range of possibilities. In individuals with GH axis defects (GHAD; DiMeglio et al., 2000) resulting from defective GH-1, pituitary transcription factor 1 (Pit-1), or GH-releasing hormone receptor genes, all present with severe congenital GH deficiency and elevated IGF-I levels. Note further that, not only will the phenotype differ depending on the alteration present in the GH-1 gene, but that individuals presenting splice site mutations, as opposed to missense mutations, may also present other pituitary hormone deficiencies. The latter deficiencies appear to occur in later postnatal stages and mainly affect the adrenocorticotrophic hormone and thyroid-stimulating hormone axes but not the gonadotrophic axis (Mullis et al., 2005).

In growth hormone insensitivity syndrome (Laron Syndrome) the GHR gene is defective. The lack of GHRs results in reduced GHBP, IGF-I, IGFBP-3 and acid labile subunit levels and an increase in serum GH (Laron et al., 1989; Guevara-Aguirre, 1996; Savage et al., 2005). The major impact of GH deficiency was believed to be in the postnatal period (Woods et al., 1996), but recent work has documented its prenatal effects (Daughaday & Harvey, 1995; Harvey, 1995; Bennett & Robinson, 2000: electronic Appendix S2). Children with mutations in the GH or GHR gene are normal or nearly so at birth with the major impairment occurring in skeletal growth (Daughaday & Harvey, 1995; Bennett & Robinson, 2000). Initially normal or near-normal sized infants soon show a difference of 5.0–11.0 SD below the mean height for age (Guevara-Aguirre, 1996; Woods et al., 1996). Body proportions with relatively short extremities remain appropriate for height, but not for age. The cranium is near normal in size but facial features remain undeveloped, giving an appearance of excessive head size (Daughaday & Harvey, 1995). Laron et al. (1989) and Merimee & Laron (1996) observed that heterozygosity for the inactivating GHR (Laron) allele results in moderately decreased GHBP levels and variable degrees of short stature.

Recently, Morwood et al. (2005, p. 1016) have suggested that abnormal growth does not explain the body proportions of LB1 as ‘…growth-hormone-related dwarfism and microcephaly in modern humans result in normal limb and pelvic proportions’. In this instance these authors are referring to a single allelic variant of an inactivating mutation of the growth hormone-releasing hormone receptor gene (GHRH-R, OMIM 139191: Maheshwari et al., 1998). This specific autosomal recessive condition, Dwarfism of Sindh, appears to have arisen in multiple villages with consanguineous couples within the last c. 30 years. Relevant clinical features include: (1) high penetrance; (2) low serum levels of IGF-I, IGF-II and IGFBP-3; (3) no response to GHRH; (4) severe postnatal growth failure (females, c. 113.5 cm; males, 118.0–136.0 cm); (5) proportionate dwarfism; (6) OFC of 3.7–6.0 SDS below controls; (7) minimal or no facial hypoplasia and (8) normal intelligence. A significant degree of truncal visceral obesity was not observed, although it was thought that a mild degree was probably present. Note that a reduction in subcutaneous fat and lean muscle mass occur in cases of GH deficiency or GH resistance whereas there is an accumulation of truncal visceral fat (Hoffman et al., 1995; Jørgensen et al., 1996; Boot et al., 1997). Maheshwari et al. (1998) also note that afflicted individuals have produced viable offspring that also express this condition. Whereas afflicted individuals do not present limb and cranial features similar to those of LB1–LB9, this example does provide a modern analogue for extreme dwarfing and brain size reductions that can result in individuals with only slightly reduced capacities consequent on a single genetic change in the GH–IGF-I axis. Also note that afflicted individuals comprise 15–20% of the population of these villages, even though there are an unexpectedly low number of affected females.

In contrast, for disease states characterized by elevated IGF-I levels the secretion of GH is impaired (Harvey, 1995). The only known human case of a homozygous partial deletion of the IGF-I gene presented with severe intrauterine growth retardation, severe postnatal growth failure, sensorineural deafness, and mental retardation (Woods et al., 1996). Of interest is that this c. 16.0-year old individual had achieved a height of 119.1 cm, a weight of 23 kg, and had long arms and microcephaly (Guevara-Aguirre, 1996; Woods et al., 1996). Individuals possessing an IGF-I deficiency or (partial) IGF-I insensitivity present with severely compromised intrauterine growth. The latter effect contrasts with the slightly diminished intrauterine growth of patients with either congenital complete GH deficiency or GH resistance, suggesting that, as in mice, the in utero production of IGF-I is largely independent of GH (electronic Appendix S2). Postnatally, IGF-I deficiency or insensitivity produces a persistent and severe growth retardation that is approximately equal in magnitude to that observed in complete GH deficiency or GH resistance. Walenkamp et al. (2005) note that this implies that postnatal growth is the result of GH-dependent IGF-I production. Of interest is that the severe growth restriction accompanying a homozygous partial deletion of the IGF-I gene contrasts with the subtle but statistically significant inhibition of intrauterine and postnatal statural and cranial growth in IGF-I haploinsufficiency (Walenkamp et al., 2005).

Walenkamp et al. (2005) observe that, as in IGF-I knock-out mice, biologically active IGF-I is necessary for normal intrauterine growth, brain and inner ear development, and mandibular growth. As IGF-I expression increases towards normal levels the impact on these craniofacial regions changes in a step-wise fashion. In cases of partial IGF-I insensitivity the functional effects on the brain are milder than seen in the homozygote, although the heterozygotes exhibit a severely reduced OFC (−4.6 SDS). In the latter case, however, the inner ear was not affected. These results suggest to Walenkamp et al. (2005) that heterozygous and compound heterozygous mutations in IGF1R do not completely preclude IGF-I signalling. Further, they indicate that the dysmorphic features seen in the different cases of IGF-I deficiency and IGF-I insensitivity may be partially due to a tissue-specific expression of IGF1R alleles (electronic Appendix S2).

GH–IGF-I axis and modern pygmies

The presence of short stature and an apparent dysfunction of the GH–IGF-I axis unites individuals with the developmental abnormalities discussed above and normal pygmy or pygmoid populations in Africa and Southeast Asia (Merimee et al., 1982, 1990; Schwartz et al., 1987; Jain et al., 1998; Clavano-Harding et al., 1999). Dávila et al. (2002) suggest that the pygmy phenotype is most consistent with resistance to the growth-promoting actions of GH. Such a resistance in pygmies is suggested by: (1) the low serum GHBP levels and, by implication GHRs; (2) the low serum IGF-I levels in adults (excepting the Mountain Ok people) and (3) a failure of short-term GH administration to induce the expected rise in IGF-I levels (Jain et al., 1998; Dávila et al., 2002). Further, Laron et al. (1989) and Merimee & Laron (1996) noted the similarity of the pygmy phenotype to that of heterozygous carriers of the Laron GHR allele (=pygmy trait). Characterization of a restriction fragment length polymorphism of GHR genes in two pygmy populations was considered a potential source of the disruption, but it is now known that the GHR gene is normal in all pygmy populations (Merimee et al., 1989: Jain et al., 1998).

Merimee et al. (1990) observed in an ontogenetic sample of pygmies that GHBP levels were less than controls for each age group (2.9–30.0 years) studied and that the per cent decrease from normal levels increased with age. These authors suggest that this evidence is most consistent with an alteration of the transcription or translation of receptors rather than with a primary defect in the GHR gene. Alternatively, it has been suggested that the main change in pygmies, excepting the Mountain Ok people, lies not in the GHR/GHBP complex (Dávila et al., 2002) but in IGF-I resistance (Jain et al., 1998). Multiple lines of evidence from investigations of Efe pygmies demonstrate that immortalized lymphocytes are resistant to the in vitro growth-promoting action of IGF-I directly, and to hormones whose growth-promoting actions are mediated by local IGF-I levels (Jain et al., 1998). The Efe also show reduced IGFR expression. Jain et al. (1998) suggest that these factors are related to short stature in the Efe. However, Jain et al. (1998) did not find a: (1) distribution abnormality of the IGF-I allele in pygmies relative to African controls; (2) defective IGF-I gene in pygmies; or (3) structural abnormality of the coding region of the IGFR1 gene (Jain et al., 1998; Dávila et al., 2002).

Whereas IGF-I resistance has been demonstrated in Efe cell lines, several features of such a resistance are problematical (Jain et al., 1998; Dávila et al., 2002; Abuzzahab et al., 2003). In cases of IGF-I resistance, it would be expected that serum levels of the peptide would increase, as opposed to the decrease characterizing most pygmy populations (Jain et al., 1998; Dávila et al., 2002). Furthermore, severe malformations arise during development in homozygous IgfR1 knock-out mice that result in death of the fetuses, although heterozygous IgfR knock-out mice are phenotypically normal (Dávila et al., 2002). Given the above, Dávila et al. (2002) suggest that the GHRs are a more likely source of the dysfunction in human pygmies. In this case, however, we do not see the expected rise in GH, as normal serum levels of the hormone are characteristic of all pygmies. Furthermore, pygmies, generally, do not respond with increased IGF-I levels when subjected to GH treatment, and their insulin concentrations do not increase (Merimee et al., 1982). If pygmies were simply resistant to GH, they should show elevated GH levels (Jain et al., 1998) and depressed IGF-I and IGF-II levels. As discussed above, pygmies have decreased levels of IGF-I and generally show normal IGF-II levels (Merimee et al., 1982; Merimee & Rimoin, 1986).

In relation to H. floresiensis, Brown et al. (2004) and Lahr & Foley (2004) reference the long-superseded results of Merimee et al. (1987), who suggested that all the effects of reduced serum IGF-I levels on growth occur during puberty in pygmies. This conclusion, based on the apparent lack of a pubertal growth spurt in African pygmies, is based on individuals with estimated ages. Bailey's (1990, 1991) work with known-aged Efe pygmies (Birth – 5.0 years) demonstrates that they are significantly shorter than controls from birth and continue to fall behind in successive age stages. It is now widely recognized that the suppression of growth in pygmy populations does not occur solely at puberty but begins prenatally.

To further complicate understanding of the GH–IGF-I axis in pygmies, both human and animal models indicate that insufficient dietary protein is associated with combined GH and IGF-I resistance (Jain et al., 1998). Whereas their tropical forest habitats predict low protein intake in pygmies, it is clear that the low IGF-I levels are related to genetic factors and not to environmental ones such as diet (Hiernaux, 1977; Merimee & Rimoin, 1986; Dietz et al., 1989). Also, Hiernaux (1977) notes that weight is more sensitive to malnutrition than stature. Given these data, Dávila et al. (2002) suggest that it appears likely that both genetic and nutritional factors contribute to short stature in pygmy populations. Therefore, the genetic background and nutritional status of pygmies may converge on the same physiological mechanism(s) to bring about short stature.

Modern pygmy populations are united in having normal serum GH levels, reduced GHR/GHBPs, low serum IGF-I levels (excepting the Mountain Ok), and sufficient protein intake to eliminate nutritional factors as the only source of these anomalous levels. To date, errors in the coding sequence for stature-related growth factors and hormones have not been found in these groups. Of interest, however, is that Foucher et al. (2002) observed that transgenic mice expressing the homeobox gene Hoxa5, under the control of Hoxb2 regulatory elements, present a growth arrest during weeks two and three of postnatal development, resulting in a proportionate Laron-type dwarfism. These mice present a liver phenotype illustrated by a 12-fold increase in liver IGFBP1 and a 50% decrease in liver IGF-I mRNA correlated with a 50% decrease in circulating IGF-I. These authors propose that growth regulation by Hoxa5 is due to its interaction in the liver with Forkhead box transcription factors. These transcription factors are key regulators of several liver genes playing important roles in postnatal growth. In particular, the Forkhead transcription factors FKHR and HNF3β/Foxa2 are considered as two essential regulators of Igfbp1 and Igf-I expression, respectively. Further, these transcription factors, including the human liver-expressed FKHR (FOXO1a), are thought to play a major role in Igfbp1 expression, especially in response to hormonal and nutritional status (Foucher et al., 2002).

Brain size evolution in primates

A general correlation exists across taxa between the degree of encephalization and cognitive complexity (Lumsden, 1983; Evans et al., 2004). Given this relationship, investigators have utilized cerebral cortex size as a proxy for cognitive abilities in both extant and extinct species (Mochida & Walsh, 2001; Evans et al., 2004). Determinations of brain size in anthropoids demonstrate that a significant increase has occurred in humans and that this mainly reflects size increase of the cerebral cortex. Rakic (1995) observes, however, that whereas the cerebral cortex has undergone different rates of expansion since the Old World monkey (macaque) - human divergence there has not been an equivalent increase in cortical thickness between these two taxa.

As discussed above, cortical neurons arise through two processes, symmetric and asymmetric mitosis of precursor cells. The duration of symmetric mitosis determines the number of radial units in the cortex of a given species and, indirectly, the size of the cortex (Rakic, 1995). Alternatively, the duration of asymmetric mitosis regulates only the number of neurons within each ontogenetic column. These observations led to the hypothesis that surface expansion of the neocortex in primates could be attributed, in the main, to a change in genetic mechanisms controlling the timing or mode of cell division, or both, and the switch between the two phases of cerebral development around embryonic day 40 (Rakic, 1995). Whereas brain size increase could result from other factors, Lumsden (1983) notes that, as the number of elements in the primary repertory increases, the amount of neural tissue enlarges unless phylogenetic transitions occur to more compact modes of packing neuronal groups into the cerebral cortex. Based on evidence of the organization of primate cortical columns there is, however, little support for the hypothesis that such a transition occurred in the primate clade (Lumsden, 1983). Further, Lumsden (1983) suggests that, as hominid brains evolved greater intellectual capacity, they would be predicted to contain more neuronal groups and thus be larger on average than the ancestral form.

Given that primate brain sizes do, in fact, appear to reflect their relative numbers of neurons, the discovery that the small cerebral cortex size of MCPH individuals results from a reduction in the number of neural precursor cells led to an immediate assessment of the rate of evolutionary change in these genes within mammalian lineages, generally, and primates, specifically. To date, only microcephalin and ASPM have been subjected to evolutionary analysis. Both of these genes show evolutionary parallels in that the rate of protein evolution for both is dramatically accelerated in the lineage leading to humans relative to that leading to other nonhuman primates and mammals (Wang & Su, 2004). Both ASPM and microcephalin have been shown to exhibit profiles consistent with portions of the gene being under strong purifying selection, whereas other portions exhibit the effects of positive selection (Zhang, 2003; Evans et al., 2004, 2005; Kouprina et al., 2004, 2005; Wang & Su, 2004; Mekel-Bobrov et al., 2005). These observations indicate that positive selection likely operated within specific functional motifs of these genes (Kouprina et al., 2004; Wang & Su, 2004). For both genes, however, the accelerated evolution is not confined to the human lineage following the human–chimpanzee divergence. For ASPM, the most dramatic acceleration occurred in the lineage from ape ancestors to humans (c. 7–8 mya: Zhang, 2003; Kouprina et al., 2004). For microcephalin, evolutionary acceleration took place in the lineage from simian progenitors to humans and chimpanzees, with the greatest acceleration occurring early (Evans et al., 2004; Wang & Su, 2004).

A direct correlation between these two genes and the massive size increase of the cerebral cortex in humans relative to chimpanzees has not been found, although recent analysis indicates that both microcephalin and ASPM have undergone recent positive selection in humans (Evans et al., 2005; Mekel-Bobrov et al., 2005). Recent microarray data also demonstrate an acceleration of gene-expression change during human brain evolution relative to that of chimpanzees (Preuss et al., 2004). These authors found three times more significant gene expression changes in the human brain relative to chimpanzees, whereas changes in the liver accumulated at comparable rates in the two lineages. Human brains also show the presence of a large number of upregulated genes when compared with chimpanzees. From these data, Preuss et al. (2004) conclude that the extensive gene-expression upregulation in humans is not a general feature of human–chimpanzee microarray comparisons but is a brain-specific phenomenon. Such an observation is consistent with an assessment of a broad array of nervous system-related genes by Dorus et al. (2004). The latter authors conclude that brain evolution in primates, particularly humans, is partially the resultant of a larger number of mutations in the coding regions of a large series of genes, especially those having developmentally biased functions. They further argue that the observed acceleration in protein evolution in a large cohort of nervous system genes represents a significant genetic correlate with the profound changes in brain size and complexity that occurred during primate evolution, especially along the lineage leading to humans (Dorus et al., 2004).

The above data clearly demonstrate that selection for brain size increase in hominids has occurred over a significant temporal span and involved a large number of genes related to neural development and function. This focus on an increase in size and support functions in an organ that is metabolically expensive, using about 20% of the body's metabolic energy, clearly indicates that factors overriding these costs do exist (Beals et al., 1984; Gilbert et al., 2005). Further, a significant reduction in brain size has never been documented in any hominid lineage, except that proposed for H. floresiensis. Brown et al. (2004) suggest that brain size reduction in this species is a product of insular dwarfing. They also note that island conditions can result in unusual disjunctions in anatomical units or functional matrices and point to evolutionary parallels in other insular mammals.

In the case of LB1, or H. floresiensis generally, it is possible that brain size reduction resulted from a single genetic change, such as those produced by GH, IGF-I, or the MCPH gene family, as opposed to a change involving a significant number of nervous system-related genes. Whereas single genetic mutations in these instances can result in a rapid change in brain size, some are already known to create genome instability, and they may provide a foundation for undesirable downstream impacts (electronic Appendices S1 & S2). Brain size is only slightly reduced on average in geographically widespread populations of modern pygmies who inhabit environments similar to the new species. This combination of stature reduction and near-normal brain size among human pygmies is produced by the ‘pygmy modification’ of the GH–IGF-I axis. Therefore, to achieve both the stature and brain size reduction seen in H. floresiensis, one must suggest either a different pathway than that selected for in all known pygmy populations, a more significant disruption of the GH–IGF-I axis, or some combination of these changes. In their study of brain–body size relationships in hominids, Schoenemann & Allen (2006) reach a similar conclusion.


With recovery of the LB1–LB9 remains, the discoverers were faced with a unique combination of morphological features. A diminutive body and brain size in combination with the lack of a developed chin in these individuals certainly suggested a link to early hominids. The presence of a suite of morphological features found in H. sapiens, a relatively recent geological context, and an association with H. sapiens-grade culture suggested more recent affinities. However, the presence of superficial similarities to early hominids and apes has become the focus of comparison for both Brown et al. (2004) and Morwood et al. (2005). In the species description, H. floresiensis was compared with early hominids and a general H. sapiens sample rather than to small-bodied modern humans or short-statured populations from prehistoric contexts on Flores and surrounding islands.

Based on the initial comparisons, Brown et al. (2004) ultimately suggest a new species that they prefer to derive from a H. erectus (s.l.) or earlier ancestor. In their scenario, a H. erectus or earlier group became isolated on the island of Flores and then: (1) either acquired or maintained the physiological changes necessary for a dwarfing condition similar to that found in modern human pygmies; (2) acquired a reduction in brain size significantly greater than the degree of stature reduction; (3) acquired a suite of H. sapiens skeleto-dental characteristics and (4) independently acquired the skills and innovation necessary to produce a stone tool technology similar in all respects to that of H. sapiens. This phylogenetic link to H. erectus, or to any taxon other than H. sapiens, results in a problem wherein the similarities between H. floresiensis and H. sapiens would have to arise by parallel evolution at the same time that the new species is undergoing significant genetic drift or selection for a reduction in stature and brain size.

If the remains attributed to H. floresiensis derive from, or are a variant of H. sapiens, how does one account for the suite of ‘primitive’ features described by Brown et al. (2004) and Morwood et al. (2005) to suggest a link with H. erectus and other primitive hominids? Morphological similarity of anatomical units or functional matrices (Moss, 1997a,b, 1997c,d) between two organisms does not a priori provide evidence of feature or character state polarity. Similarity of features or character states may result from a distant shared ancestry, may result from a recent shared ancestry, or may be due to convergent or parallel evolution.

A pertinent example of homoplasy is the physiologically induced ‘insular dwarfing’ (s.l.) in geographically dispersed human pygmy populations and its ontogenetic consequences, which result from parallel evolution rather than from recent shared ancestry (Hiernaux, 1977; Cavalli-Sforza, 1986a; Shea & Bailey, 1996). Further, reference to descriptions by Genet-Varcin (1951), Sauter & Adé (1953), Marquer (1972) and Vallois & Marquer (1976) demonstrates that features of the modern pygmy skeleton have also been described both as ‘primitive’ and ‘Sinanthropus-like’ (=H. erectus; s.s.). Some of these features and cranial regions are the same as those described in H. floresiensis by Brown et al. (2004) as being similar to H. erectus.

Weber et al. (2005) recently pointed to the gross resemblance of modern microcephalic (s.l.) brains with those of both H. floresiensis and other hominids. Because cranial vault shape is directly related to brain size and shape, Taylor & DiBennardo (1980) noted that a scaphoid-shaped skull, as described in H. floresiensis, has variously been regarded as a primitive character in H. erectus and a pathological or anomalous trait among microcephalics (s.l.). Reference to the seminal works of Vogt (1866) on microcephalics or ‘ape-men’ and the companion work by Weidenreich (1941) on craniofacial relations associated with changes in brain size provides insights into how such changes in brain size and craniofacial growth patterns in humans can result in morphologies that mimic ‘primitive’ shapes. It should be noted, however, that the perceived shape correspondence between the cranial vault and brain of H. sapiens with that of other species of Homo or of apes is dependent on the degree of brain size reduction. Given this fact, a phylogenetic link between H. floresiensis and H. erectus based on their shared possession of a ‘primitive’ cranial vault and brain endocast shape (Antón, in Falk et al., 2005a,b; Wong, 2005; Morwood et al., 2005) cannot be validated. Furthermore, the cranial capacity of the extremely platycephalic cranium of H. floresiensis is at least 50% smaller than that typically found in Indonesian H. erectus. Because of this fact, the cranial vault of H. floresiensis cannot possess a ‘H. erectus shape’ (s.l. or s.s.), especially in combination with the disjunction in brain size reduction relative to body size reduction.

The suggested primitive features of the LB1 skull and LB6 mandible could represent independently derived results of evolutionary or anomalous or pathological developmental processes operating on a H. sapiens substrate. Because it is probable that modern pygmy populations have recent origins (Bailey & DeVore, 1989), it is likely that at least some of the so-called ‘primitive’ features they have been described as sharing are, in fact, uniquely or secondarily derived homoplasies. Given such a scenario, some of the skull features described as primitive in H. floresiensis may be linked to body size reduction and craniofacial changes resulting from modifications in the GH–IGF-I axis, as in modern pygmies, and to brain size reduction. However, modern pygmies do not share the extreme degree of brain size reduction as seen in H. floresiensis. Because of this difference, features that appear to be ‘primitive’ in pygmies and that might be considered to be related to a reduced brain size cannot be so linked. These features are most probably related to GH–IGF-I axis modifications.

Further work is needed to document the craniofacial anatomy of pygmy populations and to determine how this anatomy relates both to their reduced stature and to the skeletal anatomy of the LB1–LB9 remains. Only with reference to the concept of functional cranial components, pioneered by de Beer (1985), or functional matrices as championed by Moss (1997a,b, 1997c,d), will a morphological assessment reveal the underlying cascade of effects on the included matrices. In summary, it is currently not possible to differentiate the so-called ‘primitive’ features of the skull observed in modern pygmy populations and individuals with reduced brain sizes from those described for H. floresiensis, or to determine how either of these feature sets relates to those described for H. erectus (s.s.).

Regardless of the taxonomic and phylogenetic implications, it appears biologically reasonable that the basic craniofacial skeleton of H. floresiensis could have been derived from a H. sapiens template. Is this true for the ‘primitive’ features of the post-cranial skeleton and body size? As discussed above, the lower-end of the stature range in modern human pygmy females very closely approximates the estimated statures of the LB1 and LB8 individuals (Table 1). Morphological features of the skeleton (wide pelvis, long arms relative to legs, tibial cross-sectional shape, etc.) that are said to link H. floresiensis with early hominids are also found in modern human pygmy populations. Some of these features have been described as ‘primitive’ in pygmies and most are linked to body size reduction. Shea & Bailey (1996) caution that researchers who have characterized such post-cranial features of modern pygmies as divergent or primitive have failed to appreciate the ways in which a basic modification of mechanisms controlling body size can ramify along multiple allometric pathways to produce seemingly quite different terminal morphologies. Discussions of the ‘primitive’ nature of the LB1 ilium need to include observations from pygmies and individuals with GH deficiencies or insensitivities. In the latter cases, it is well known that there is an accumulation of truncal visceral fat, and this may have an impact on the orientation of the ilium, especially if the locomotor behaviour was unusual. Of the features noted for H. floresiensis, the only ones that do not fit the general predictions derived from observation of modern human pygmy skeletal remains are the long bone circumference-to-length ratios. The LB1 and LB8 individuals are clearly unusual in this respect, and this difference has important physiological, behavioural and locomotor implications. Unfortunately, available data are insufficient to determine the significance of these differences. In summary, the post-cranial features known for H. floresiensis should not be used in taxonomic or phylogenetic assessments without first making extensive comparisons with short-statured modern populations and resolving issues related to the long bone shaft-to-length ratios.

Many of the post-cranial features described for the Flores remains (Brown et al., 2004; Morwood et al., 2005) are consistent with those found in human populations that have undergone stature reduction by way of a modification of the GH–IGF-I axis. Brain size in humans can be reduced through the MCPH gene family or changes in IGFs and GHs (via GHRH-R inactivation) to the size of that found in H. floresiensis without significant impacts on the viability of a majority of the affected individuals. Genetic mutations that produce such reductions can be present at high frequencies in small, consanguineous groups in relatively isolated contexts. MCPH individuals show developmental delay and, in some, a reduction in stature. However, despite considerable effort I have been unable to uncover data on how MCPH impacts somatic development and, by extension, how this might impact craniofacial development. I have also been unable to discern how changes resulting from MCPH might interact with developmental modifications expressed in GH–IGF-I axis defects. Nevertheless, based on available data, a basic model that accounts for the morphology of H. floresiensis can be constructed.

In this scenario, a H. sapiens population on Flores would acquire a dwarfing condition in combination with a mutation of the MCPH gene family. In this context, it is important to recognize that there is no reason to assume that a single insult accounts for every morphological feature of the Flores remains. There are no known direct impacts of modification or defects in the GH–IGF-I axis or in the MCPH gene family on other coding or regulatory regions. Once a GH–IGF-I/MCPH combination arose, it would be expected that a significant number of individuals within the population would have been affected, and that there would be significant phenotypic variation. As discussed above, it is known that some rural populations have an c. 20–45% occurrence rate of microcephaly and that individuals in MCPH families only show a deviation in OFC from one another of ±1.0 SD. Given these data, I would expect multiple MCPH individuals to have been present in a single locality and the group to have shown a range of brain sizes.

Such a scenario would account for the reduced body and brain size found in the LB1–LB9 individuals, but it does not fully reveal the underlying biology of the craniofacial skeleton. Any attempt to reveal this biology and construct a basic explanatory model is complicated by the lack of: (1) individuals possessing a combined GH–IGF-I/MCPH malformation; (2) documentation on specific changes and geographic variation in skeleto-dental features in individuals possessing either of these conditions and (3) knowledge of the change that results in the modification of the GH–IGF-I axis in pygmy populations.

We currently lack ‘direct’ evidence (i.e. a skeleton) of the impact of a combination of MCPH and the ‘pygmy modification’ of the GH–IGF-I axis. Having such an individual would be of interest, but it would still only represent one possible combination of an array of potential phenotypes. The practice of relying on a single malformation or syndrome to account for the observed morphology or on a single malformed individual for comparative purposes (Brown et al., 2004; Falk et al., 2005a; Morwood et al., 2005) already plagues analysis of this material, and is simplistic.

Complications arise in assessing available data for the new ‘species’ as there is an absence of condition-specific and geographical data on craniofacial changes for either aspect of the GH–IGF-I/MCPH combination. Based on available data, the LB1 skull appears to show a disparity in functional cranial components similar to that found in humans with significantly reduced brain sizes or microcephaly (s.l.: Weidenreich, 1941; Richards, 1985). The interpretation of these modifications is complicated by the fact that descriptions of individuals with a reduction in brain size similar in scale to MCPH individuals derive mainly from Caucasian populations, with a single African adult and Mongoloid-affiliated infant known (Vogt, 1866; Weidenreich, 1941; Richards, 1985). The actual condition affecting these microcephalics (s.l.) is, in most cases, unknown. Whereas Suzuki (1975) provides some limited insight into geographic variation of the craniofacial region in microcephaly (s.l.), data for Southeast Asian and Australomelanesian groups are particularly lacking. The overall complexity of the task we face in modelling craniofacial changes under these conditions can be gleaned from recent assessments of Cohen syndrome. Studies by Hennies et al. (2004) and Mochida et al. (2004) demonstrate the degree of allelic heterogeneity and phenotypic effect present in an ethnically diverse group afflicted with this single syndrome.

Further, because of the lack of specific craniofacial data for MCPH or microcephalic (s.l.) individuals, it is currently difficult to model the impact on the anterior mandible in an individual possessing a GH–IGF-I axis modification. Given the morphology of some pygmy mandibles, it appears that only minor effects on maxillo-mandibular ontogeny would be needed to produce the observed morphology of LB1 and LB6. Such a morphological suite could derive from the combined effect of physiological changes and the range of modifications in positioning of functional matrices that characterize microcephalics (s.l.). Although I do not consider the mandibular configuration of available microcephalics (s.l.) to resemble that of H. floresiensis, disruption of the included functional matrices within the suggested scenario might produce the observed configuration. It should be noted, however, that a lack of direct evidence does not imply that the ontogenetic pathway is inconceivable, only that the exact genetic–physiological pathway and its precise morphological consequences are difficult to predict.

Given these data and the known variation in the microcephalic (s.l.) phenotype, the statement by Antón (in Wong, 2005, p. 57) regarding the dissimilarity of facial patterns between H. floresiensis and ‘microcephalics’ can be put into an appropriate context. Such statements underscore the general lack of understanding that a range of phenotypes are expressed in any given pathological or anomalous condition and that this variation is compounded by ontogenetic and geographic variation. Whereas we lack geographic and other specific data on craniofacial patterns associated with significantly reduced brain sizes, we are fortunate to have numerous geographically dispersed populations that share a similar change in the GH–IGF-I axis. These populations possess geographic variation in craniofacial and post-cranial proportions that should prove useful in evaluating the impact of these physiological changes.

We are currently limited by not knowing the exact mechanism(s) that produces dwarfing in modern pygmies or the specific aspects of craniofacial ontogeny affected. As discussed above, the physiological changes in modern groups may be related to GH, IGF-I, or their respective receptors or binding proteins. Further, complications arise in that both aspects of the axis: (1) have autocrine and paracrine functions in addition to the better known endocrine function of GH; (2) possess complex interactions with their receptors and binding proteins which can limit or enhance their availability for specific tissues or functions and (3) vary significantly in their effects during ontogeny (electronic Appendix S2). Also, there is no guarantee that the Flores individuals share the same physiological pathway to dwarfing as modern pygmies. I have described relevant aspects of the GH–IGF-I axis and how changes in this axis significantly but differentially affect ontogenetic processes. I believe that it is this complexity and the potential detrimental impacts of even minor changes to the system that have seemingly eliminated other possible mutations of the GH–IGF-I axis in modern populations and resulted in the near homogeneity of change found in pygmy populations worldwide. Clearly other pathways are possible and have arisen over time; the Flores remains may represent one of these divergent pathways. Most such alternate pathways appear, however, to have undesirable downstream impacts that have resulted in their rapid or relatively rapid elimination from the human gene pool.

In summary, whereas multiple models can be constructed to account for the morphology of the LB1 and LB6 remains, the simplest is the combined effects of MCPH and a ‘pygmy modification’ of the GH–IGF-I axis (Table 2). The next most likely model is simply one in which an alteration occurs in the GH–IGF-I axis. As discussed, it is known that a complete lack of GHRH-Rs and, by extension, a lack of GH generally can result in both small brain size and reduced stature. It is also known that a reduction in IGF-I during development results in a similar effect on neurogenesis, disruption of the cell cycle, to that caused by members of the MCPH gene family (electronic Appendix S2). Further, it is known that a reduction in IGF-I during ontogeny results in brain size reduction, mandibular modifications and significant reductions in body size. Because of the broad spectrum of impacts on ontogeny that result from various modifications of the GH and IGF-I systems, it is not a simple task to delimit specific morphologies. Further, it is clear that alterations in the IGF-I pathway have the potential to carry with them significant and undesirable impacts on ontogeny in general (Denley et al., 2005; Woelfle et al., 2005). For this reason, as a hypothesis to explain the unusual morphology of the Flores hominid remains, I suggest the combination of MCPH and a modification of the GH–IGF-I axis similar to that in modern pygmies as a likely scenario (Table 2). Although probably not necessary, I would also suggest a slightly earlier impact on IGF-I production during ontogeny than that found in modern pygmies. It should be noted, however, that no matter how one approaches the phylogeny of this ‘species’, the genetic and physiological changes that resulted in the observable morphology most certainly followed similar pathways to those outlined here.

Table 2.   Timing of onset during ontogeny of the combined effects of normal and abnormal microcephalin, ASPM, GH–IGF-I axis, and the known or projected phenotype
ConditionOntogenetic stageGeneralized phenotype
Early embryonicLate embryonicPrenatalPostnatal
  1. *The use of the term ‘pygmy trait’ in Merimee et al. (1982) refers to the situation of having a low serum concentration of IGF-I but normal IGF-II levels. These authors consider this expression as similar to the general situation in pygmy populations. My usage of the term ‘pygmy’ modification is more generalized as pygmy populations differ in how they achieve short stature and differ slightly in their expression of the GH–IGF-I axis.

  2. This is referring to the ‘general pygmy’ condition of normal GH levels, reduced GHR/GHBPs, low serum levels of IGF-I, and normal serum levels of IGF-II.

  3. Refer to the text and Woods et al. (1996) for a discussion of the potential effects of an IGF-1 gene deletion.

Possesses an MCPH gene; GH–IGF-I axis normalDisruption of cell cycle and apoptosis regulation during neurogenesis (microcephalin) or disruption of the modulation of mitotic spindle activity in neuronal progenitors (ASPM)Variable but significantly reduced cerebral cortex; Potential for slightly reduced somatic growthBrain and cranial vault significantly reduced in size; Potential for slightly reduced somatic growthBrain growth essentially complete with little subsequent increase; Cranial vault small whereas other skull components show decreasing effect relative to their distance from the brain; Impact on somatic growth ranges from mild to severeVariable but significantly reduced brain size (≥4–14 SD), but brain structure is normal; Mental retardation (mild to severe) but no other neurological defects; Stature ranges from normal to significantly reduced but impact is less than that on brain size
Lacks MCPH genes; GH portion of the GH–IGF-I axis is abnormal Neurogenic impact ranges from none to severe; Visuomotor disturbances in severe casesSignificant to little impact on brain development; Somatic development is slowed but near normalImpact on brain may be severe, with associated visuomotor disturbances and impaired mental development, or it may be only slight; Near normal cranium but underdeveloped face; Somatic development now significantly impacted, at 5–11 SD below mean for ageBroad range of impacts associated with component disrupted (GH, GHR, GHBP; GHRH-R); Brain may be relatively normal or significantly impacted, with reduced size; May have mental retardation and visuomotor defects; Stature variable but significantly below mean for age and sex; Possible other, but unrecognized, effects on neural and somatic tissues
Lacks MCPH genes; IGF-I portion of the GH–IGF-I axis is abnormalOnly partial deletions and insensitivity syndromes known and result in mild to severe impacts on intrauterine development; Virtually all body regions are affected; Lack of IGF-I may have a similar neurogenic effect as MCPHCascade of effects starting with brain, inner ear, and then mandible; Number of impacts is reduced in reverse order with decreased severity; Major impact on somatic growthViable infants have significantly reduced but variable brain sizes and may have sensorineural deafness and reduced mandibular growth; Somatic growth significantly impactedBrain development may be significantly or severely impacted (>4.5 SD minimum); Cranial vault small with skull components showing a decreasing effect relative to their distance from the brain; Additional but variable impacts on the skull in the temporal and mandibular regions; Somatic development significantly reducedSignificant range of impacts depending on location of malformation (IGF-I, IGF1R, IGFBP); Complete absence results in fetal death whereas insensitivity syndromes show postnatal signs similar to GH related conditions; Brain severely to moderately impacted; May have sensorineural deafness and mandibular growth defects; Early impact on somatic growth that continues into postnatal period
Microcephalin/ASPM gene + GH–IGF-I axis abnormal but lacks ‘pygmy’ modification*Disruption of normal neurogenesis with decreased numbers of neuroprogenitor cells; Probable early and unmoderated effects of reduced levels of IGF-I, its carriers, and/or its receptors Variable but significantly reduced cerebral cortex; Probable early, unmoderated effects on brain and somatic development from reduced levels of GH and IGF-I and their carriers, and/or receptorsSignificantly reduced brain size and stature with probable increased and unmoderated effects on stature at puberty due to reduced levels of GH and IGF-I and their carriers, and/or receptorsBrain reduced significantly with mild to severe mental retardation; Stature significantly below mean for age and sex; Increased phenotypic effects due to an inability to moderate the effects of the disruption of the GH–IGF-I axis
Microcephalin/ASPM gene + GH–IGF-I axis with ‘pygmy’ modificationDisruption of normal neurogenesis with decreased numbers of neuroprogenitor cells; Probable early but moderated effects of reduced levels of IGF-I, its carriers, and/or its receptors Variable but significantly reduced cerebral cortex; Moderated effects on brain development; Demonstrated effects of reduced levels of IGF-I, its carriers, and/or its receptors on somatic developmentSignificantly reduced brain size and stature with probable increased but moderated effects on stature at puberty of reduced levels of IGF-I, its carriers, and/or its receptorsBrain reduced significantly with mild to severe mental retardation; Stature variably reduced but significantly below mean for age and sex; Potential for abnormalities in both brain and somatic development due to disruption of the GH–IGF-I axis; Overall effects moderated via mutation of the axis pathway, as found in modern pygmies


I find no support for a generic level distinction or an assignment to Australopithecus or H. habilis for the Liang Bua Cave, Flores, Indonesia remains that derive from Pleistocene deposits. The derivation of H. floresiensis from a H. erectus ancestor is, at present, equally unsupported. Brown et al. (2004) believe that their recovery of a small series of stone tools demonstrates that H. erectus inhabited Flores 840 kyr ago and could, therefore, provide the ancestral population (Morwood et al., 1998). There is, however, a c. 750 kyr interval between this suggested occurrence and the oldest of the Liang Bua Cave remains, an isolated radius shaft.

The presence of these multiple individuals in the Liang Bua depository has been claimed by the discovery team to strengthen the hypothesis that LB1 is not an isolated pathological individual but is a member of a long-term population. For example, whereas Brown et al. (2004) did not assign the oldest remains (the radius shaft) to H. floresiensis, this element has now been so assigned (Morwood et al., 2005). Note, however, that the length estimated from this shaft is consistent with that of modern pygmies (Henneberg & Thorne, 2004). The significance of the recovery of lower third premolar teeth from three individuals, which have been described as different from H. sapiens, is also an open question. The isolated LB2 premolar potentially indicates populational continuity between this individual and the later LB1 and LB6 individuals but does not indicate shared brain and body sizes. These additional individuals appear to show that LB1 is not an isolated pathological individual (contra Antón, in Wong, 2005, p. 57; Henneberg & Thorne, 2004; Jacob, 2004), but additional and more complete remains are obviously needed to assess the extent of variation among individuals and through archaeological time. Only additional excavation will illuminate those dimensions.

As the isolated radius shaft is not diagnostic and the significance of the isolated lower premolar (LB2) is unclear this would place the totality of diagnostic remains assigned to H. floresiensis within the time range of c. 12–18 kyr. It is also thought that H. sapiens had populated surrounding islands by at least c. 60 kyr ago (Morwood et al., 2004; Macaulay et al., 2005; Thangaraj et al., 2005) and may have also inhabited Flores by this time. As discussed above, remains of H. sapiens are known from deposits directly above those containing H. floresiensis. Given these circumstances, the only support for linking the LB1–LB9 remains to H. erectus are the perceived primitive features of the skull and post-cranial skeleton. I have discussed how many of these ‘primitive’ features may be related to: (1) brain and body size reduction; (2) disruption of the GH–IGF-I axis, as observed in modern pygmies and (3) geographic variation in craniofacial morphology.

Morwood et al. (2005) conclude that the H. floresiensis remains do not represent pathological or aberrant individuals but represent a long-term population that exhibits a morphological suite never recorded in normal or pathological H. sapiens. I agree that these remains do not represent diseased, pathological, or aberrant individuals. I consider these individuals to manifest physiological differences from other modern humans in similar ways to modern pygmies and additional modern humans possessing MCPH or GHRH-R mutations. When developmental differences result in divergent morphologies that then define a population, they can provide the basis for speciation, but by themselves they do not confer specific status. The fact that the Liang Bua remains present with a currently distinct suite of characters relative to modern humans does not necessarily mean that they represent a separate species. We currently lack the kind of detailed morphological assessments of both pygmy populations and individuals possessing the kinds of anomalous conditions suggested to account for the observed morphology. Prior to suggesting that the observed morphological suite of the LB1–LB9 individuals is indicative of speciation, it will be necessary to collect a substantial amount of new data.

When one applies the concept of maximum parsimony (Sober, 1988) to the totality of evidence available on the Flores remains, one finds significant support for the remains being a variant of H. sapiens and little support for a species-level distinction. Given this position, I suggest that the LB1–LB9 individuals represent the remains of a H. sapiens group which became dwarfed in an island environment via changes in the GH–IGF-I axis. Acquisition of a dwarfing condition may either have occurred prior to or after the group arrived on the island. If it can be demonstrated that the totality of the recovered remains sample the same population, it appears that a mutation in the MCPH gene family or a secondary modification of the GH–IGF-I axis arose in the later part of their occupation of the island and was transmitted within a local group. Whereas I consider the ‘primitive’ features identified in the LB1–LB9 individuals to be consistent with the scenario presented above, only a detailed analysis will be able to clarify the value of these features for phylogeny reconstruction.

The acquisition of such a set of features in an insular context is certainly one way in which populations diverge and give rise to new species. The brain size reduction observed in the LB1 individual, if applicable to all contemporaneous individuals on Flores, differs from the trend in all known members of the genus Homo. It may have provided a selective advantage or been a neutral adaptation for this group. In an island context lacking other H. sapiens, a brain size reduction produced by mutation of MCPH related genes or GH–IGF-I axis modifications could conceivably have provided a beneficial reduction in metabolic costs without a detrimental reduction in intellectual capacity. However, based on modern populations where similar brain size anomalies have arisen, a significant number of partially afflicted or nonafflicted individuals would have been present. Given the known emphasis on brain size and upregulation of brain related gene expression in Homo across diverse ecogeographic contexts, the potential for such a malformation arising and resulting in a selective advantage is unlikely.

Analysis of DNA, mtDNA, and additional skeletal remains from well-controlled stratigraphic contexts will be necessary to choose among competing hypotheses and to elucidate the place of the Flores people in human evolution. Whatever the ultimate resolution of the phylogenetic and taxonomic issues stemming from the recovery of these fascinating remains, the morphological pattern they manifest must have arisen from developmental pathways. Fortunately, these pathways are being elucidated by modern research and were definitively responsible for the unique morphology of the Pleistocene remains from Flores.


  • 1

    Hiernaux (1977) notes that mean stature values for ‘pygmy’ males is <150.0 cm whereas ‘pygmoids’ have heights between 150.0 and 160.0 cm (Table 1). The criterion for pygmoids is, however, based more on socio-economic factors than biological ones (Hiernaux, 1977). Given this situation the term ‘pygmy’ is used herein to refer to populations with a mean male stature of ≤160.0 cm.


This work benefited from comments by and conversations with: Rebecca S. Jabbour (New York Consortium in Evolutionary Primatology); Drs Tim D. White, F. Clark Howell, Denis Su and Jean-Renaud Boisserie (Human Evolution Research Center, UC Berkeley) and Dr Dorothy Burk (University of the Pacific, Arthur Dugoni School of Dentistry, San Francisco). I also thank Drs Tim D. White and F. Clark Howell (HERC) and Dr Dorothy Dechant (Institute of Dental History and Craniofacial Study, University of the Pacific, Arthur Dugoni School of Dentistry, San Francisco) for access to specimens in their care.

Notes added in proof

Martin et al. (2006) have recently responded to Falk et al. (2006), presenting new data they interpret as indicating that the brain of the Flores holotype individual is too small for its body size, given predictions on brain-body size relationships in mammals. These authors also suggest that these remains may represent a pathological individual with microcephaly.

Falk, D., Hildebolt, C., Smith, K., Morwood, M.J., Sutikna, T., Jatmiko, Saptomo, E.W., Brunsden, B. & Prior, F. 2006. Response to Comment on ‘The brain of Homo floresiensis’. Nature312: 999c.

Martin, R.D., MacLarnon, A.M., Phillips, J.L., Dussubieux, L, Willians, P.R. & Dobyns, W.B. 2006. Comment on ‘The brain of Homo floresiensis’. Nature312: 999b.