Flores hominid: New species or microcephalic dwarf?



The proposed new hominid “Homo floresiensis” is based on specimens from cave deposits on the Indonesian island Flores. The primary evidence, dated at ∼ 18,000 y, is a skull and partial skeleton of a very small but dentally adult individual (LB1). Incomplete specimens are attributed to eight additional individuals. Stone tools at the site are also attributed to H. floresiensis. The discoverers interpreted H. floresiensis as an insular dwarf derived from Homo erectus, but others see LB1 as a small-bodied microcephalic Homo sapiens. Study of virtual endocasts, including LB1 and a European microcephalic, purportedly excluded microcephaly, but reconsideration reveals several problems. The cranial capacity of LB1 (∼ 400 cc) is smaller than in any other known hominid < 3.5 Ma and is far too small to derive from Homo erectus by normal dwarfing. By contrast, some associated tools were generated with a prepared-core technique previously unknown for H. erectus, including bladelets otherwise associated exclusively with H. sapiens. The single European microcephalic skull used in comparing virtual endocasts was particularly unsuitable. The specimen was a cast, not the original skull (traced to Stuttgart), from a 10-year-old child with massive pathology. Moreover, the calotte does not fit well with the rest of the cast, probably being a later addition of unknown history. Consideration of various forms of human microcephaly and of two adult specimens indicates that LB1 could well be a microcephalic Homo sapiens. This is the most likely explanation for the incongruous association of a small-brained recent hominid with advanced stone tools. Anat Rec Part A, 288A:1123–1145, 2006. © 2006 Wiley-Liss, Inc.

Brown et al. (2004) recently recognized a new hominid species, “Homo floresiensis,” on the basis of skeletal remains recovered from the limestone cave of Liang Bua on the Indonesian island of Flores. In a companion paper, Morwood et al. (2004) reported on associated stone tools and faunal remains, providing dates bracketed between 38,000 and 18,000 years ago for the relevant sediments. The time depth for H. floresiensis and associated stone artifacts was extended to 74,000–95,000 years ago by Morwood et al. (2005a). The primary specimen (LB1), from an uppermost level dated at about 18,000 years ago, is an associated skull and partial skeleton from a dentally adult individual. The most immediately striking feature of the LB1 skeleton is its small size. Maximum length of the femur is 280 mm, slightly less than the minimum value of 281 mm recorded for Australopithecus afarensis (AL-288-1) and equal to the minimal estimate for the Homo habilis skeleton (OH62). Taking maximum femur length, stature of the LB1 skeleton was estimated at 106 cm using formulae derived from human pygmies (Jungers, 1988), and a body mass of 16.0–28.7 kg was then inferred from this stature. (For comparative purposes below, a mid-range value of 23 kg is taken.) One key feature, which gives the visual impression of primitive morphology in LB1, is the absence of a chin in the mandible. An even more striking, and certainly unexpected, feature of the skull of the main specimen is its very small cranial capacity. Brown et al. (2004) reported a value of only 380 cc measured with mustard seed. Indeed, because of the small cranial height associated with the small brain size of this individual, Brown et al. (2004) stated that their inferred stature of 106 cm was likely to be an overestimation. Despite this very small cranial capacity, a follow-up study of a virtual endocast derived from the LB1 skull concluded that the brain shows a number of similarities to that of Homo and is closest to that of Homo erectus (Falk et al., 2005a).

In addition to the main skeleton LB1, fragments of two other individuals were reported in the initial publications. Brown et al. (2004) referred to Homo floresiensis an isolated mandibular premolar (left P3) dating back at least 37.7 kyr and stated that “additional evidence of a small-bodied adult hominin is provided by an unassociated left radius shaft, without the articular ends, from an older section of the deposit (74–95 kyr).” Morwood et al. (2004) gave a body height estimate of about 1 m based on that radius shaft. Additional elements of the LB1 skeleton and further remains attributed to six additional individuals were subsequently reviewed by Morwood et al. (2005a). The only substantial new specimens reported are a second mandible (LB6) that resembles the LB1 mandible in lacking a chin and a second right tibia (LB8).

The overall conclusion initially derived from all of the skeletal material from an inferred total of nine individuals is that Homo floresiensis was a dwarf form derived from Homo erectus. Dwarfing was interpreted as a result of isolation on the island of Flores, paralleling known cases of evolutionary dwarfing in certain island-living mammals (e.g., elephants). In fact, dwarf elephants (genus Stegodon) are found in the same deposits, and it was concluded that Homo floresiensis not only made the stone tools found in the deposits but also hunted juvenile Stegodon and possibly even used fire for cooking, in view of the presence of charred animal bones (Morwood et al., 2005a). This scenario was reiterated in a popular account by Morwood et al. (2005b). Among other things, it has been suggested that the Flores hominid shows that a large brain size is not a prerequisite for the production of advanced stone tools (Wong, 2005). Given such far-reaching interpretations, the finds have naturally attracted considerable attention.

Because the Flores finds involve such a strikingly incongruous combination of a tiny-brained hominid with advanced stone tools, it is surely advisable to give serious consideration to alternative explanations (Henneberg and Thorne, 2004; Jacob et al., 2006; Martin et al., 2006; Richards, 2006). One possibility is that the individual represented by the main skeleton LB1 suffered some kind of pathology, exhibiting a form of microcephaly. For the time being, only a single skull is known for Homo floresiensis and that skull happens to have a very small cranial capacity. In this sense, it is undoubtedly microcephalic, i.e., small-headed (see also Richards, 2006). It is important to note at once that small brain size and small body size are to some extent separable issues. It is perfectly possible that Flores was inhabited by a small-bodied hominid species between 100,000 and 18,000 years ago. It is well established that human body size tends to decrease with decreasing latitude, accompanied by increasing average annual temperature, and is particularly small in warm, humid climates at low latitudes (Roberts, 1953; Ruff, 1994; Katzmarzyk and Leonard, 1998). Furthermore, a skeleton from an adult human “pygmoid” with an estimated age of about 30 years and a height of only 146 cm has been reported from the separate cave site of Liang Toge on Flores (Verhoeven, 1958; Jacob, 1967). Richards (2006) provides an extensive discussion of reduced stature in various modern human pygmy populations. LB1 could be a pathological, tiny-brained individual from such a population of very short stature.

Rejection of the possibility of pathological microcephaly in LB1 by Brown et al. (2004) was seemingly supported by a subsequent publication by Falk et al. (2005a), which compared virtual endocasts of LB1, modern human, Homo erectus, chimpanzee, and “a European microcephalic.” It was concluded that the endocast of LB1 was distinctly different from the single microcephalic included in that comparison. Later on, however, a study of 19 human microcephalic skulls revealed considerable variation in external brain morphology, with the endocast of one individual seemingly showing overall similarity to that of LB1 (Weber et al., 2005), although that conclusion was questioned (Falk et al., 2005a).


As Brown et al. (2004) themselves noted, the value of 380 cc they reported for the cranial capacity of LB1 is comparable to the lowest values recorded for Australopithecus and lies well within the range for great apes (e.g., chimpanzees). In fact, the volume of the virtual endocast subsequently reported for LB1 by Falk et al. (2005a) is appreciably larger, at 417 cc. The discrepancy of 37 cc between these two reported values for cranial capacity, almost 10%, is disconcertingly large. Nevertheless, even the higher value of 417 cc is strikingly small in comparison to all other known hominids. The initially reported value (380 cc) is smaller than in any other known undoubted hominid apart from two individual Australopithecus afarensis (343 cc in AL 333-105; 375 cc in AL 162-28), although the higher value of 417 cc also exceeds that reported for more recent Paranthropus aethiopicus (410 cc for WT 17000). In fact, the volume of a computer-generated endocast of LB1 has since been reported to be 400 cc (Holloway et al., 2006), and that value (close to the mean of the two previously reported values) will be taken for purposes of further discussion here. Accordingly, it is necessary to go back about 3.5 million years (my) to find a hominid cranial capacity as small as that of the Flores hominid (Fig. 1). All values reported for the Flores hominid (380–417 cc) are also smaller than in most gorillas and fall well within the range for common chimpanzees (Fig. 2). Hence, it is unquestionable that brain size in the LB1 skeleton of Homo floresiensis, dating back only 18,000 years, was tiny by any standard.

Figure 1.

Cranial capacities recorded for 118 fossil hominids plotted against time, extending back almost 3.5 Ma (data from Stanyon et al., 1993). The arrow indicates the highly incongruous value (red circle) reported for Homo floresiensis at only 18,000 years ago. The recently reported values for the four Dmanisi skulls (magenta circles) fall well within the range for hominids dated at around 1.7 mya and fit the general trend, in striking contrast to the value for H. floresiensis.

Figure 2.

Histograms showing cranial capacities for African great apes (Gorilla gorilla; n = 48; Pan troglodytes; n = 95) and various fossil hominids (Australopithecus spp., n = 10; Paranthropus spp., n = 8; Homo habilis, n = 7; Homo erectus, n = 28; archaic Homo, n = 17; Homo neanderthalensis, n = 22; Homo sapiens, n = 26). Data for the African great apes are taken from the records of Adolph Schultz (Anthropological Institute, University of Zurich); data for fossil hominids are from Stanyon et al. (1993). The pink vertical bar indicates the range covered by the reported values for Homo floresiensis (380–417 cc).

Figure 3.

Comparison of the original microcephalic skull in Stuttgart (left; Staatliches Museum für Naturkunde, 5297/25523) with the cast held in the collections of the American Museum of Natural History, New York (right; AMNH No. 2792a). Note the clear difference in coloration of the calotte compared to the rest of the skull in the AMNH cast.

At first sight, it might be thought that the diminutive cranial capacity of LB1 could be attributable to evolutionary dwarfism, as suggested by Brown et al. (2004), although Argue et al. (2006) note that “insular dwarfism is unknown for Homo to date.” However, it is well established that reduction of body size within a mammal species (including Homo sapiens) is usually associated with only moderate reduction in brain size. Whereas the exponent value for scaling of brain mass size to body mass in comprehensive interspecific comparisons across placental mammals is close to 0.75 (Martin, 1981; Martin et al., 2005), there is a progressive decline with decreasing taxonomic rank and the value for intraspecific scaling among adults of a single species is typically about 0.25 (Martin and Harvey, 1985; Kruska, 2005). One of the best-documented cases is that of the domestic dog, with an exponent value of 0.27 determined for 26 breeds covering a 21-fold range of body sizes generated by artificial selection (Bronson, 1979). In fact, in modern humans and other primates examined, the exponent value is generally lower than 0.25 and approaches zero if males and females are considered separately (Martin and Harvey, 1985). Even with an intraspecific scaling exponent value of 0.25, halving of body mass would only be expected to result in reduction of brain mass to 84% of its original value. In modern human pygmies, for example, cranial capacity is not greatly different from that in populations of larger body size. Falk et al. (2005a) note that human pygmy skulls typically have cranial capacities exceeding 1,000 cc (compared to a worldwide mean value for all modern humans of 1,349 cc) (Beals et al., 1984). The adult female pygmy skull used in their comparison in fact had a cranial capacity of 1,249 cc, while modern Rampasasa pygmies on Flores have an average cranial capacity of 1,270 cc (Jacob et al., 2006), similar to the value of 1,204 cc reported by Jacob (1967) for the Flores “pygmoid” described by Verhoeven (1958).

Brown et al. (2004) explicitly suggested that Homo floresiensis was derived from Homo erectus through a process of insular dwarfing (see also Morwood et al., 2005b), although Morwood et al. (2005a) state that “H. floresiensis is not just an allometrically scaled-down version of H. erectus.” Examination of this proposal is complicated by considerable divergence in the definition of Homo erectus. Here, a very broad view with no geographical restriction will be taken, as this effectively covers all options for comparison. In one general survey using such a broad definition (Stanyon et al., 1993), mean cranial capacity for 28 Homo erectus was 991 cc (range, 727–1,251 cc). If brain size scales to body size with an exponent value of 0.25, the body size of Homo erectus would have to be reduced to one-eighth of the original value for a cranial capacity of 400 cc to be included at the lower end of the range. In fact, for a cranial capacity of 400 cc to correspond to the mean value of dwarfed Homo erectus, body size would have to be reduced to 1/32 of its original value (Martin et al., 2006). In other words, an original body size of 60 kg for Homo erectus (Kappelman, 1996) would have to be reduced to just 2 kg for the mean cranial capacity to be reduced from 991 cc to an average value of 420 cc. A more recent survey of cranial capacity in broadly defined Homo erectus, with an increased sample size of 38 (Krantz, 1995), reported a higher mean value of 1,045 cc (range, 780–1,360 cc). This would correspond to an even greater body mass reduction required to attain the cranial capacity of LB1 in a dwarf form.

It could be argued that some individuals included in the broadly defined taxon Homo erectus have quite small brains, and that the cranial capacity reported for Homo floresiensis would be more likely to result from dwarfing of such small-brained individuals. A case in point is provided by four skulls from the Dmanisi deposits in Georgia, dated at about 1.7 mya. Although these specimens have in fact been referred to the taxon Homo ergaster, for geographical reasons they could conceivably be relevant to the origin of the Flores hominid. The four skulls from Dmanisi have a mean cranial capacity of only 664 cc (range, 600–775 cc) (Gabunia et al., 2000; Vekua et al., 2002; Rightmire et al., 2006). It has been suggested that the Flores hominid descended from such a small-brained population, but this proposal is also unconvincing. First, with an intraspecific scaling exponent of 0.25, body size would still have to be reduced to one-eighth of the average for Homo erectus (from 60 to 7.5 kg) to decrease mean cranial capacity from 664 to 400 cc. In fact, taking the smaller body mass of 50 kg estimated for the Dmanisi fossils (Gabunia et al., 2001), body size would have to be reduced to around 6 kg to reach a mean cranial capacity of 400 cc. Contrary to the impression that has been given, cranial capacities of the Dmanisi hominids are not unusually small, given their antiquity. The four values recorded fall within the range previously found for hominids at 1.7 mya (Fig. 1). Indeed, the antiquity of the Dmanisi hominids renders any direct comparison with the Flores hominid inappropriate because cranial capacity shows a general trend to increase over time within the broadly defined taxon Homo erectus and among hominids generally. Taking the sample of 28 values for Homo erectus given by Stanyon et al. (1993), a trend line indicates that average cranial capacity increased by about 200 cc over the period covered. As Homo floresiensis is only 18,000 years old and has been identified as a remarkably late-surviving evolutionary dwarf form of Homo erectus, it seems more likely that dwarfing would have taken place from a relatively large-brained late representative. It should also be noted that in one of the cranial comparisons conducted by Argue et al. (2006), including a single skull from Dmanisi (D2280), there was no indication of any morphometric affinity with the LB1 skull.

In both temporal and geographical terms, the representatives of the taxon Homo erectus that are closest to the Flores hominid are the Ngandong specimens from the Solo River in Java. Dating of those specimens has been subject to much uncertainty. They were originally thought to date back around 200,000 years or more. However, preliminary radiometric dating indicated an age of 50,000–100,000 years (Bartstra et al., 1988), and subsequent dating using a combination of radiometric measurement and electron spin resonance yielded an even younger age of 27,000–53,000 years (Swisher et al., 1996). Hence, the Ngandong specimens may possibly be only 9,000–35,000 years older than the LB1 skeleton. The average cranial capacity for six skulls from Ngandong is 1,149 cc (Stanyon et al., 1993), almost three times larger than that of the Flores hominid.

This all leads to the conclusion that it is simply unrealistic to explain the tiny cranial capacity of 380–417 cc recorded for Homo floresiensis as an outcome of evolutionary dwarfism affecting an insular population of late-surviving Homo erectus (Martin et al., 2006).

Several instances of evolutionary dwarfism in mammalian lineages are known from the fossil record as well as from recent species, ranging from squirrels and sloths to hippopotami and mammoths. Pleistocene dwarf elephants, for example, are known from a number of Mediterranean islands. The presumed ancestral mainland species, Elephas antiquus (Caloi et al., 1996), had an average estimated body mass of 10,000–15,000 kg, while the comparatively tiny island dwarf form, E. falconeri, from Malta-Sicily, had an estimated mass of only 100 kg (Roth, 1992). The difference in brain size between these two species was much less marked than the hundred-fold difference in body size. The cranial capacity of E. antiquus was approximately 9,000 cc, whereas that of E. falconeri was 1,800 cc (Accordi and Palombo, 1971). The slope of the line joining these two sets of values corresponds to an exponent value of 0.32–0.35, much closer to the typical intraspecific scaling value of 0.25 than to the interspecific scaling value of about 0.75 for mammals generally. Brain size dwarfing in elephants therefore resembles the scaling pattern determined across wide body size ranges for domestic dogs (scaling exponent = 0.27) (Bronson, 1979), sheep (0.29) (Bronson, 1979), horses (0.25) (Bronson, 1979), and wild boar (0.24) (Kruska, 1970). The time taken for the dwarfing of Mediterranean elephants is unknown, but could be up to several hundred thousand years (Ambrosetti, 1968; Caloi et al., 1996; Lister, 1996). However, the dwarfing process can occur very quickly. This is shown by the case of red deer on the island of Jersey, which became dwarfed to about one-sixth of their original size (estimated body mass of about 36 kg for dwarf adult males compared with about 200 kg for the ancestral mainland species) over a period of 6000–11,000 years (Lister, 1996).

A special case explicitly invoked by Brown et al. (2004), Brown and Morwood (2004), and Argue et al. (2006) to account for the remarkably small cranial capacity of LB1 is a report on unexpectedly small brain size in a dwarfed insular bovid by Köhler and Moyà-Solà (2004). It was found that six chronologically successive species of the extinct genus Myotragus found in Pliocene and Pleistocene deposits of Majorca had relatively small brains compared to other living and fossil bovids. Because the inferred sister genus (the Pliocene rupicaprine Gallogoral) resembles the other bovids in its relative brain size, it was concluded that a marked decrease in relative brain size (by about 50%) had taken place following the isolation of Myotragus on Majorca by 5.2 mya. Despite the apparent parallels to the case of the Flores hominid, however, there are crucial differences. Most importantly, investigation of relative brain size in Myotragus was initiated because of the strikingly small size of the orbits, suggesting marked reduction in size of the eyes. No such reduction in orbit size has been suggested for the Flores hominid, and there is indeed no evidence thereof. Furthermore, the time scale concerned is one or two orders of magnitude greater and involves a distinction at the generic level (between Myotragus and other bovids) rather than at the intraspecific level (if LB1 is interpreted as a dwarfed Homo erectus). In fact, a comparative study of a short DNA sequence extracted from Myotragus (Lalueza-Fox et al., 2002) indicates that this genus is not closely related to rupicaprines after all (hence ruling out Gallogoral as the sister genus) and is most closely related to Ovis, which was not included in the comparative sample studied by Köhler and Moyà-Solà (2004). The ancestry of Myotragus is at present undetermined, so we have no direct information about relative brain size at the outset. Inferred reduction in relative brain size in Myotragus thus has little or no relevance to the tiny brain of the Flores hominid.


Some scholars believe that the archaeological history of the Island of Flores begins during the Lower Pleistocene/Middle Pleistocene transition, based on equivocal evidence from the site of Mata Menge (dated to ca. 800,000 BP). At that site, 14 stone tools out of a total of 54 artifacts were identified originally (Morwood et al., 1998, 1999), while an additional 507 artifacts were found in recent excavations (Brumm et al., 2006). The objects concerned were found in river gravels in association with a Stegodon, although there is some question whether all of them are actually artifacts, rather than accidents of nature.

Brumm et al. (2006) suggest that there are “similarities, and apparent technological continuity” between the flakes produced at Mata Mange and Liang Bua, the site where remains of H. floresiensis was recovered. There are, however, real questions concerning the association of the artifacts with the fission track dates because Brumm et al. (2006) mention “hydraulic transportation and size sorting.” Furthermore, the suggestion that there is cultural continuity over a period of almost 800,000 years is quite surprising and represents a view of lithic technology that is at odds with our understanding of production and use of stone tools. Because flakes were produced at Mata Menge does not mean they represent a “tradition” (see Clarke, 1968). Instead, they may be just flakes, as one would find in any assemblage at any time and anywhere in prehistory. Additionally, the few cores from Mata Menge are not of types ever found with an Acheulean or Mousterian assemblage, but are recorded from the Upper Paleolithic period onward. Further, Brumm et al. (2006: Fig. 4) consider the relationship between two perforators found at Liang Bua and several pieces they claim are the same type from Mata Menge. In this instance, we have to consider the fact that perforators have never been recovered from Acheulean or Mousterian sites on any continent where these technocomplexes have been found. If the flakes and tools from Mate Menge do indeed have any relationship with those recovered at Liang Bua, it is far more likely that that they are from the same time period (ca. 18,000 BP) and must have been transported either by the movement of Homo sapiens or by water.

Figure 4.

Age distribution for 68 cases of human primary microcephaly surveyed by Hofman (1984). Note that most cases (ca. 54%) are from individuals that died before exceeding the age of 21 years. Data set from Hofman (1984), kindly provided by the author. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

The presumption of Morwood and colleagues is that the maker of the Mate Menge implements was Homo erectus, remains of which had been found on the island of Java early in the 20th century, but never on Flores. Although artifacts of Middle Pleistocene age are attributed to fossil hominid finds on Java (Bartstra, 1992; Keates and Bartstra, 2000), the exact association is equivocal (Corvinus, 2004), and there is some question of whether some are artifacts at all, whether on Java or Flores. Thus, stone artifact production earlier than 30,000 or so on Flores, and therefore the presence of Homo erectus, is not confirmed.

The belief stated by Morwood and colleagues is that Homo erectus remained isolated on Flores for the remainder of the Pleistocene. It is proposed that, along with other mammals, Homo erectus became progressively smaller, until dwarf mammals and giant reptiles (e.g., Komodo dragons) were the main animal species on Flores ca. 18,000 BP (but see Allen, 1991). Morwood et al. (2004) also hypothesize that their supposed dwarf form of Homo erectus developed hunting practices—together with evolved artifacts—that replicate artifact assemblages (and, presumably, hunting patterns) of modern Homo sapiens in other parts of Indonesia (e.g., East Timor) (O'Connor et al., 2002). This leads on to a particularly incongruous feature of the proposed interpretation of the recently discovered Flores material from Liang Bua cave: All of the stone tools (n = 32) reported from the level of section VII containing the LB1 skeleton by Morwood et al. (2004), as well as those described from section IV (which are even more advanced than those in section VII), clearly belong to types that are consistently associated with Homo sapiens and have not previously been associated with H. erectus or any other early hominid. The artifact assemblage clearly involved a tradition using the prepared-core technique, which is confined to Homo neanderthalensis and Homo sapiens. Furthermore, the explicitly noted presence of bladelets is a hallmark of Homo sapiens, found in Africa, Asia, Europe, and after ca. 40,000 BP. Yet Morwood et al. (2004) concluded that “H. floresiensis made the associated stone artifacts.”

In fact, two anomalous features of the section VII assemblage are evident in Figure 5 of Morwood et al. (2004): 5g is a Levallois core (not a burin core for producing microblades; the small bladelet-like removals are features of the core preparation); and 5c, the bipolar core, like the Levallois core, produced flakes, not blade or bladelet blanks. In other words, the blanks in the assemblage are blades and bladelets, while the cores correspond to production of flakes. The most likely interpretation is that these blade and bladelet blanks, although found near the cores, are not actually associated with them. For section IV, stratigraphically partially above the section VII material and located toward the center of the cave, the authors mention the finding of thousands of artifacts, “up to 5,500 per cubic meter,” mainly flakes produced from radial cores (i.e., Levallois). However, they also mention a “more formal component … including … blades and microblades,” thus reinforcing the interpretation that there are at least two sets of assemblages in the cave, both seemingly associated with their hominid.

Figure 5.

Histograms showing brain mass in adult microcephalics (n = 16 female; n = 17 male), with an overall average vale of 427 g. Data set from Hofman (1984), kindly provided by the author. [Color figure can be viewed in the online issue, which is available at www. interscience.wiley.com.]

It is inherently unlikely that the reported complex of Upper Paleolithic blanks and tools would have been developed independently by an unusually small-brained dwarf evolutionary descendant of H. erectus. Any alternative explanation invoking secondary acquisition by Homo floresiensis of tools or tool-making techniques from Homo sapiens would raise a host of additional unanswered questions. The normal expectation would be that any hominid dated at 18,000 years ago associated with tools typical of Homo sapiens would be a member of that species. The anomalies in the archaeological data most likely indicate the presence of fully modern competent Homo sapiens utilizing the Liang Bua cave many times after their arrival on Flores.


Given that the brain size is so unusually small in the LB1 hominid despite its remarkable young geological age (Fig. 1), the possibility of a pathological disorder, specifically some form of microcephaly, must be considered (Henneberg and Thorne, 2004; Jacob et al., 2006; Martin et al., 2006; Richards, 2006). This possibility was mentioned briefly in the original report (Brown et al., 2004), but then rejected. The main authors of the original reports on the LB1 skeleton also defended this position in dismissing the proposal by Henneberg and Thorne (2004) that LB1 might be a pathological specimen (Brown and Morwood, 2004). That proposal was explicitly addressed in a subsequent paper comparing a virtual endocast of the LB1 skull with endocasts from an adult female chimpanzee, an adult female H. erectus (specimen ZKD XI from Zhoukoudian), a modern human female, and a single “European microcephalic” (Falk et al., 2005a). The origin of the European microcephalic skull was not otherwise described, but the authors reported that “its shape conforms to features of its corresponding skull that typify primary microcephaly (microcephalia vera): small cranial vault relative to face, sloping forehead, and pointed vertex.”

On closer examination, the single microcephalic skull of unstated provenance taken by Falk et al. (2005a) for their comparison has proved to be an inappropriate choice for several reasons (Martin et al., 2006; Richards, 2006). The published account provides no details of the skull that was investigated. It was particularly important to establish whether the skull concerned was that of an adult as in the case of LB1 or derived from an immature individual suffering from a pronounced pathology resulting in early death. Another cause for concern is that the cranial capacity of this microcephalic skull, which is not explicitly stated by the authors, is exceedingly small. Although its remarkably small size is not obvious from the illustrations provided by Falk et al. (2005a), which were all scaled to a standardized volume (rather than scaled in proportion), it is evident from the diminutive linear dimensions provided in the accompanying table.

In response to our enquiry, Falk reported that the microcephalic specimen examined had been obtained from the collections of the American Museum of Natural History (AMNH) in New York. Falk also noted that the specimen's skull shape typifies that associated with microcephalia vera and stated that the specimen had anomalous teeth. For that reason, the specimen's age at death was not estimated, although it was believed to be a juvenile. In fact, it is simply impossible to take any single skull as typical of “true microcephalics.” The term “primary microcephaly” (microcephalia vera) is a general descriptor applied to individuals that have an unusually small brain size at birth, recently defined as −3 standard deviations at birth (Dobyns, 2002) or −4 standard deviations at older ages (Woods et al., 2005), reflecting impairment of brain development attributable to a great variety of syndromes (Gilbert et al., 2005). As noted by Mochida and Walsh (2001), “the condition is clearly genetically and clinically heterogeneous.” The key point is that the microcephalic skull examined by Falk et al. (2005a) was used for a comparison with adult representatives of all other taxa despite the fact that it was thought to come from an immature individual. Although brain size reaches adult dimensions early in development (typically 6–7 years in normally developing humans), early death of the microcephalic individual studied renders direct comparison with the adult LB1 inappropriate.

Direct measurement of the cranial capacity of the AMNH specimen using glass beads yielded a cranial capacity of only 260 cc (K. Mowbray, personal communication). Hence, the apparent brain size of this specimen was remarkably small not only in comparison to that of LB1 (about 65% of the value) but also in comparison to the usual range of modern human microcephalics. In fact, the published images of the microcephalic endocast in the supplementary data provided by Falk et al. (2005a) also exhibit several unusual features that are not seen in humans with the more common forms of microcephaly. Few gyral indentations are apparent in comparison to all of the other endocasts, and the frontal pole is pointed. The cerebellum is small, although it is not as drastically reduced as the cerebrum. Also, both the occipital lobes and the cerebellum appear to hang down further and at a sharper angle than observed in most hominid brains. Finally, the foramen magnum is seemingly greatly enlarged, a defect known to be associated with cerebellar malformations. These rare anomalies, combined with the extremely small cranial volume, suggest that the individual suffered from a severe brain malformation and not isolated or primary microcephaly.

Examination of the microcephalic skull studied by Falk et al. (2005a), housed in the collections of the AMNH (Fig. 5), revealed the surprising fact that it is not an original specimen but a plaster-based cast. It bears the accession number 2792a and is contained in a box bearing an inscription indicating that it came from Cannstatt, Germany. However, the catalogue entry states “Casts of microcephalus skull from Plattenhardt and fragments from Cannstatt, Germany (Wurtemberg).” The cast itself bears the inscription “Plattenhardt. Tausch mit Stuttgart 1907” (“Plattenhardt. Exchange with Stuttgart 1907”). Available records provide no further information concerning this specimen, other than the fact that it was included in a large collection purchased from Felix von Luschan by the AMNH in 1924. The teeth on the cast (eight in the upper jaws and nine in the mandible) are highly unusual. They are small, widely separated, and peg-like, with apparent signs of heavy wear on the crowns. It is indeed virtually impossible to determine a reliable age from the cast of this individual using standard dental criteria. One reasonable interpretation would be that only one molar is present, on the right side of the lower jaw, and that this is the skull of a child.

Further enquiries revealed that the original skull from which the AMNH cast had been made almost a century ago is in fact still included in the collection of the Staatliches Museum für Naturkunde, Stuttgart (Dr. Elmar Heizmann, personal communication; Fig. 3). The skull has two accession numbers: 5297 (former registration system) and 25523 (new registration system). Using standard criteria, the age of this individual at death was estimated to be 12–13 years (Dr. Doris Morike, personal communication).

A remarkable feature of the AMNH cast 2792a is that the calotte (a separate element) is cream-colored, whereas the rest of the cranium and the mandible are dark brown (Fig. 3), suggesting that the latter parts were varnished at some time. Furthermore, the calotte does not fit properly onto the rest of the skull and the profile of the cut line deviates in several respects between the two. Whereas sutures are clearly apparent on the calotte, their continuation cannot be traced down into the lower part of the cranium. These disparities, taken together, raised the distinct possibility that the calotte was not in fact part of the original cast and had been manufactured subsequently. In order to check this, small samples were taken from the two separate parts of the cranium and subjected to chemical analysis using an inductively coupled plasma-mass spectrometer (ICP-MS) in the Department of Anthropology at the Field Museum in Chicago. It emerged that there are striking chemical differences between the two parts (Table 1). The level of calcium, effectively serving as a control, is virtually identical in the two samples, but there are major differences in other elements. The level of lead is approximately 50 times higher in the lower part of the skull (compatible with the interpretation that there had been previous treatment with a lead-based varnish), whereas the calotte shows markedly higher levels of manganese (2.7×), barium (3.5×), lanthium (14.6×), and cerium (14.4×). By contrast, tin—a major component in the elemental profile—is three times higher in the lower part of the skull cast than in the calotte. Other notable differences in that part of the cast are seen in boron (5×), sodium (2.7×), and potassium (2.4×). These major differences in elemental composition demonstrate beyond reasonable doubt that the calotte was created from a different batch of plaster and has a questionable connection with the rest of the cast. This would directly concern any impressions of gyri on the dorsal surface of the virtual endocast generated from this specimen by Falk et al. (2005a). The dangers of studying a cast instead of the original skull are further illustrated by the fact that the cross-sectional area of the foramen magnum is in fact about 17% greater in the AMNH cast than in the original skull. This increase exaggerates the marked pathological appearance of the foramen magnum and adjacent features of the endocast. Furthermore, there are discrepancies between the cast and the original skull with respect to volumetric measurements. Measurement by one of us (R.D.M.) of the cranial capacities of the original skull and the AMNH cast, using fine lead shot, yielded values of 269 and 268 cc, respectively. Although these values are almost identical, the volume of the ventral part of the cranial cavity is larger in the original skull than in the AMNH cast (139 vs. 130 cc), whereas the volume of the calotte is smaller (130 vs. 138 cc).

Table 1. Results from chemical analysis of the two parts (calotte and remaining cranium) of the cast of a microcephalic skull (AMNH 2792a). (Results generated using an inductively coupled plasma-mass spectrometer (ICP-MS) by Laure Dussubieux and P. Ryan Williams.)
  1. lod = limit of detection

  2. N.B. The value for chlorine (Cl35) must be discounted because hydrochloric acid was used to dissolve the samples.


It was also discovered that the Stuttgart skull was included in an early discussion of human primary microcephaly by Vogt (1867). That survey covered 10 skulls from Germany in considerable detail, including descriptions of endocasts for 9 of them, and also incorporated information from 19 other cases. Only 6 of the overall sample of 29 cases (i.e., 21%) involved adult individuals aged 21 years or more. In a more extensive review conducted much later by Hofman (1984), the proportion of adult individuals aged 21 years or more was higher (31 out of 68; i.e., 46%; Fig. 4), but it is nevertheless clear that many microcephalics represented in collections and surveys died before reaching adulthood. On the other hand, survival into adulthood is not uncommon, and Hofman's survey included one woman with a height of 1.34 m who survived to the age of 74 with a brain mass of only 277 g. The average cranial capacity reported for nine of the skulls examined by Vogt was 410 cc (393 cc for the four adult individuals aged 21 years or above). Comparable figures are indicated by the sample examined by Hofman (1984) (Fig. 5), with a mean brain mass of 421 g for adult females (n = 16) and a mean of 433 g for adult males (n = 17). All of these values match well with the observation from clinical experience that the brain volume of human primary microcephalics is about 400 cc, an estimate consistent with typical adult head circumferences of 40–45 cm in living human cases personally examined by one of us (W.B.D.). The full range of postnatal head circumferences in human microcephalics lies between −4 and −12 SD (Woods et al., 2005).

The skull in the collection of the Staatliches Museum für Naturkunde in Stuttgart is undoubtedly that of an individual named Jakob Moegele from the village of Plattenhardt, who died at the age of 10 years and 1 month. Vogt (1867) recorded the cranial capacity of that individual as 272 cc, the smallest value determined in his survey. Interestingly, 3 of the 10 individual skulls examined by Vogt came from closely related individuals; in addition to the skull of Jakob Moegele, he also examined that of his brother Johann Georg Moegele (who died at 5 years of age) and that of his cousin Johann Moegele (who died at 15 years of age). Jakob Moegele was the 8th of 11 children born in a single family. In addition to his elder brother Johann Georg (the 6th child), the 2nd child (also named Jakob) and the 11th child (Barbara) were likewise microcephalic, giving a total of 4 microcephalic individuals among the 11 siblings. Even more intriguingly, a total of seven microcephalic children were born to four couples (two of them unrelated to the two Moegele families) in the village of Plattenhardt within the space of a few years. This aroused sufficient attention for a special report to be commissioned from the court physician (Vogt, 1867).

The extreme pathological nature of Jakob Moegele's skull is clearly revealed by CT scans. Both the endocranial cavity and the virtual endocast (which has a calculated volume of 268 cc) show the unusual shape of the brain, with both the occipital lobes and the cerebellum hanging down conspicuously (Fig. 6). The teeth, aptly described as “mushroom-like” by Vogt (1867), are highly unusual in shape and position, and the developing replacement teeth that would be expected in a 10-year-old child are completely lacking in both upper and lower jaws (Fig. 7).

Figure 6.

Virtual reconstructions from CT scans of the right hemiskull and endocast of the Stuttgart microcephalic specimen. Note the highly unusual dentition and the downward-hanging occipital lobes and cerebellum. Images prepared by Jonathan Brown.

Figure 7.

Maximum intensity projection derived from CT scans of the left upper and lower jaws of the Stuttgart microcephalic, revealing the aberrant structure of the teeth and the complete absence of developing replacement teeth. (Image prepared by Jonathan Brown.)

In contrast to the aberrant skull of Jakob Moegele, microcephalic skulls and endocasts that are much closer in morphology to the Flores LB1 specimen most certainly do exist. One hemiskull of a dentally adult male microcephalic that is held in the collections at the Hunterian Museum in London (RCSHM/Osteo 95.1) is quite similar in size and external appearance (Fig. 8). The museum catalogue indicates that this specimen came from India, and it is described in a note by Shortt (1874) in which he states that the individual concerned was 5 ft. 6 in. tall and weighed 89 lb (40.3 kg). Further information is provided in a review of 19 microcephalic skulls by Humphry (1895), who included an illustration of the Hunterian skull (his Fig. 1 of skull 1) confirming its origin in India. Doubling of the endocranial volume measured from the Hunterian hemiskull yields a cranial capacity of 432 cc, very close to the value recorded for LB1. The anthropological collections of the Field Museum in Chicago also include plaster casts of a dentally adult human microcephalic skull (accession number A219679) and an accompanying endocast (accession number A219680). The catalogue entry indicates that the skull and endocast (Fig. 9) are from a “microcephalic idiot” from Basutoland (now Lesotho). The specimens were acquired as part of the Marshall Field Archaeological Expedition to Western Europe in 1927–1928 and entered the collections in 1931. A label attached to the endocast indicates that it was produced in England by R.F. Damon, but no further information is available in the museum records. In fact, plaster casts of the skull and mandible of this same individual are present in the collections of the American Museum of Natural History in New York (accession number 99.1 2601 A,B), and another endocast is held in the Hunterian Museum of the Royal College of Surgeons, London (RCSHM D684.4). A literature search revealed that this case had been reported by Dru-Drury (1919–1921), who stated that the individual concerned was a 32-year-old woman with severe mental retardation. She reportedly had the body size of a 12-year-old child and a body mass of only 60 lb (27.2 kg). The cause of death was recorded as tuberculosis. An endocast from this individual was later included and illustrated in a discussion of four microcephalics by Weidenreich (1941), who noted the presence of the skull cast in the AMNH collection. Weidenreich (1941: p. 396) stated that “the form and proportions of the brain … as much as can be defined from the endocast, fail to show any appreciable differences when compared with the normal human brain.” Yet the volume of the Lesotho microcephalic endocast is even smaller than that of the Hunterian specimen, amounting to only 335 cc. Both of these endocasts, illustrated in Figure 10, have a relatively normal external appearance, lacking the evident pathologies shown by the brains of some human microcephalics such as that of Jakob Moegele. The only really obvious macroscopic anomaly in both cases is the extremely small size.

Figure 8.

Comparison between a hemiskull of an adult male human microcephalic from the collection the Hunterian Museum in London (RCSHM/Osteo 95.1) and the LB1 skull (after Brown et al., 2004), both drawn to scale. Drawing by Jill Seagard.

Figure 9.

Plaster casts of a dentally adult human microcephalic skull from Lesotho and an accompanying endocast in the collections of the Field Museum, Chicago (accession numbers A219679 and A219680, respectively). (Photographs by John Weinstein).

Figure 10.

Drawings of (right) an endocast from the hemiskull of the human microcephalic from India in the collections of the Hunterian Museum (RCSHM/Osteo 95.1) and (left) of the left side of a human microcephalic endocast from Lesotho in the collections of the Field Museum, Chicago (accession number A219680). Illustrations by Jill Seagard.

A recent comparative study of virtual endocasts from 19 human microcephalics by Weber et al. (2005), with an average volume of 404 cc, has emphasized the considerable range of variability in brain shape that exists. The authors of that study singled out one particular endocast that appeared to be quite similar to that of the Flores hominid LB1, although this interpretation was questioned by Falk et al. (2005b) on the grounds that the degree of similarity was decreased if the two endocasts were oriented in the same way. Unfortunately, Weber et al. (2005) did not provide identifying information for the specimens examined and did not state which individuals were adults.

In order to avoid disagreement about endocast orientation and description (Falk et al., 2005b, 2006) and to achieve direct comparability with the results reported by Falk et al. (2005a), we repeated some of their multivariate analyses, expanding their sample by adding an endocast produced from the Hunterian microcephalic skull from India, and an endocast from the same museum of the Lesotho specimen. We also included a new endocast produced from the original Stuttgart microcephalic skull [corresponding to the virtual endocast from the AMNH skull cast used by Falk et al. (2005a)]. Two principal component analyses were carried out on the same sets of indices as those used by Falk et al. (2005a). The first set comprised six indexes (five being used in the analysis) derived from four external endocast dimensions: length, breadth, height, and frontal breadth. The eight indexes in the second set were derived from six measurements of the base of the endocasts (Table 2). In a minor departure from Falk et al. (2005a), to enhance clarity of data presentation we opted for two-dimensional plots of the first and second principal components (PC1 and PC2) rather than three-dimensional plots including PC3. Most of the information is contained in PC1 and PC2 (Table 3), and three-dimensional plots (which are difficult to interpret on the printed page) do not reveal any major differences. In a plot of PC1 against PC2 for the external endocast indices (Fig. 11a), the distribution of specimens included in the original analysis by Falk et al. (2005a) is very similar to that shown in their Figure 2A and corresponds to descriptions in their text. In Figure 11a, the endocast prepared directly from the original skull of the Stuttgart microcephalic, newly included here, falls close to the virtual endocast derived by Falk et al. (2005a) from the AMNH skull cast. The approximate matching confirms general agreement between our measurements; the minor separation between the points may reflect measurement deviations and/or distortion in the virtual endocast derived from the poor-quality AMNH skull cast. The other new specimens included in Figure 11a, the Hunterian and Lesotho microcephalic endocasts, fall close together and are not far removed from LB1. All three of these specimens are clearly separate both from the Stuttgart/AMNH microcephalic endocasts and from normal modern humans. Insofar as the complexity of brain shape can be captured by such simple indices, this demonstrates that LB1 is not clearly distinct from all modern human microcephalics and more closely resembles the two new specimens included here than the more severely pathological single specimen studied by Falk et al. (2005a).

Table 2. Endocast measurements and derived indices (following Falk et al., 2005a) for the 3 new specimens included here (Hunterian, Lesotho and Stuttgart microcephalics)
inline image
Table 3. Results of principal component analyses carried out on the samples illustrated in Figure 11, showing the first two principal components (PC1, PC2)
(a) External endocast indices
Principal componentEigenvaluePercentage of varianceCumulative percentage of variance
Principal component 1Principal component 2
(b) Basal endocast indices
Principal componentEigenvaluePercentage of varianceCumulative percentage of variance
Principal component 1Principal component 2
Figure 11.

Plots of the first two principal components (PC1, PC2) from analysis of indices derived from (a) external endocast measurements and (b) basal endocast measurements. The combined samples comprise endocasts from modern humans of average stature [Homo sapiens (1)], a modern human pygmy [Homo sapiens (2)], two australopithecines (Sts 5; WT 17000), five Homo erectus (Trinil 2; ZKD III, X, XI, XII), LB1, chimpanzees, gorillas, and the AMNH microcephalic skull cast, all from Falk et al. (2005a), with the addition of three microcephalic endocasts: the Hunterian, Lesotho, and Stuttgart (original for the AMNH cast) specimens (Tables 2 and 3).

In a plot of PC2 against PC1 for the basal endocast indices (Fig. 11b), the specimens included in the analyses of Falk et al. (2005a) are again distributed very similarly, corresponding to their Figure 2C and to the descriptions in their text. Falk et al. did not include the AMNH microcephalic endocast in this analysis. Figure 11b includes all three microcephalic endocasts in our sample: the Hunterian, Lesotho, and Stuttgart specimens. They are scattered very widely across the plot, with the Hunterian microcephalic falling closest to WT17000, Gorilla and Pan, the Lesotho microcephalic located very close to Sts 5, and the Stuttgart microcephalic closest to modern humans. (It was necessary to estimate basal measurements involving the olfactory bulbs for the Hunterian specimen because the impression on the skull from these structures is not clear. Analysis without this specimen yields a very similar distribution for the remaining specimens.) Intriguingly, as in the original plot published by Falk et al. (2005a), LB1 lies very close to Homo sapiens in Figure 11b. The overall conclusion that can be drawn from analyses of these indices is that modern human microcephalics are clearly very variable, and that specimen choice would greatly influence any analysis based on a limited sample. Figure 11b suggests that such principal component analyses of basal endocast indices are not useful for determining whether or not LB1 could be a modern human microcephalic. Analyses of external endocast indices are potentially more useful, and the results presented here support the conclusion of Weber et al. (2005) and Martin et al. (2006) that LB1 is quite similar to some modern human microcephalics.


All of the factors discussed above led us to give more detailed consideration to the possibility of pathological microcephaly raised by Henneberg and Thorne (2004), particularly as the authors of the original report on the LB1 skeleton did not discuss the relevant medical disorders known among modern humans. In the original paper, Brown et al. (2004) state without explanation “neither pituitary dwarfism, nor primordial microcephalic dwarfism in modern humans replicates the skeletal features present in LB1.” The references cited in support of this statement do not present a modern understanding of pathological conditions in modern humans characterized by severe short stature with microcephaly. We agree that pituitary dwarfism and at least one type of “primordial dwarfism,” known as Majewski osteodysplastic primordial dwarfism type 2 [MOPD type 2; taken as the focal condition by Argue et al. (2006)], differ in several respects from the LB1 skeleton, yet that skeleton clearly shares many features with syndromes of severe short stature and microcephaly as a group, a point to which we will return.

Similarly, Falk et al. (2005a) state that primary microcephaly or microcephalia vera “is characterized by small cranial vaults relative to facial skeletons, sloping foreheads and pointed vertices” and imply that the lack of these shape features in the LB1 skull excludes primary microcephaly. That conclusion is unjustified as low sloping foreheads and pointed vertices are not seen in all affected individuals with primary microcephaly (Woods et al., 2005). Falk et al. (2005a) go on to state that “microcephaly with simplified gyral pattern (MSG) is another form of congenital microcephaly … manifesting reduced numbers and shallowness of cortical sulci. The cortical topography of LB1's endocast precludes it from this form of microcephaly.” In fact, MSG is not another form of congenital microcephaly at all, only a descriptive term that one of us (W.B.D.) has used to describe the appearance of the brain in individuals with primary microcephaly (Dobyns and Barkovich, 1999; Barkovich et al., 2001). In any case, the endocasts shown in the paper by Falk et al. (2005a) lack the fine details of the gyral pattern and depths of sulci that would be needed to recognize an MSG pattern.

Thus, the analyses in both the initial paper describing the LB1 skeleton (Brown et al., 2004) and the subsequent report on the virtual endocast (Falk et al., 2005a) do not adequately reflect current understanding of human microcephaly and syndromes involving severe short stature with microcephaly. Both of these publications assume that only a few types exist, whereas a search of the OMIM database using the single search term “microcephaly” finds more than 400 genetic syndromes associated with microcephaly (see also Gilbert et al., 2005). This figure is cited by Argue et al. (2006), although Richards (2006) gives a lower figure of 300.

Any discussion of specific syndromes must rely on correct interpretation of the taxonomic status of the LB1 fossil, which remains controversial. The primary published papers (Brown et al., 2004; Morwood et al., 2005a) devote little attention to the potential existence of pathological features. Yet examination of a living modern human with similar features in a medical genetics clinic would yield the following conclusions for a young adult female with height ∼ 106 cm, body mass 16–29 kg, and head circumference (our estimate) 39–41 cm; relative to modern human standards, these values would be graphed at −9 to −10 standard deviations (SD) for height, −4 to −6 SD for body mass, and −10 to −12 SD for head circumference. Physical examination would reveal a recessed jaw with no chin, accompanied by congenital dental anomalies consisting of absent mandibular right P4 and maxillary right M3 (questioned by Jacob et al., 2006), small maxillary left M3, and pathological rotation of both maxillary P4s. [Lukacs et al. (2006) provide additional comments on the dental anomalies of LB1.] The long bones of the LB1 fossil appear disproportionately broad and less modeled (less narrowing of the diaphysis) than long bones in modern humans, as would be seen on radiographs (see also Jacob et al., 2006). All of these abnormalities taken together would lead to diagnosis of a severe short stature with microcephaly syndrome, although the available data are not sufficient to match this to a specific known syndrome (Judith G. Hall, personal communication).

If LB1 originated from a population in which very short stature was characteristic, the head size, or at least brain size, would still be too small (−5 to −6 SD as discussed below), and other syndromes including primary microcephaly would be considered in the differential diagnosis. Some of these syndromes are compatible with survival into adult life, given help from “normal” individuals. This relates back to our concerns regarding the capabilities of the extant population first raised in the section on the stone tools found in Liang Bua. Importantly, essentially all of the syndromes in the differential diagnosis have autosomal recessive inheritance and have the potential to recur within a small, inbred population. Hence, as occurred in the mid-1800s in the small village of Plattenhardt with the Moegele family, it is entirely possible that more than one individual with the same syndrome could occur in the same place, despite the overall relative rarity of the condition. Jacob et al. (2006) estimate that any human hunter/gatherer population inhabiting Flores would have been quite small, thus increasing the likelihood of inbreeding. However, for the same reason, these authors question the likelihood of survival of an isolated population on Flores for over 800,000 years without immigration.

As a next step, we can formally assess the reported dimensions of LB1 in the light of alternative interpretations regarding the fossil's population of origin. LB1 could represent a microcephalic individual from a modern human population either with normal stature (hypothesis 1a) or with dwarfed stature (hypothesis 1b). LB1 could also represent a microcephalic individual from a contemporaneous Homo erectus population (dwarfed or undwarfed, hypotheses 2a and 2b) or an early offshoot of a more primitive hominid line (hypothesis 3). Of course, in all cases the types of microcephaly to be considered must be restricted to those in which survival to adulthood is possible. This would include microcephaly with near normal cognitive abilities or mild-moderate mental retardation. Human microcephalic syndromes can be divided into two categories, a high-functioning group and a low-functioning group (Dobyns, 2002; Gilbert et al., 2005). The former category is most relevant for comparison with LB1, an individual that survived to adulthood, although in some cases survival into adulthood may occur even with moderate to severe microcephaly, as in human Seckel syndrome.

Under the hypothesis that LB1 comes from a dwarfed population derived from either Homo sapiens or Homo erectus, we assume that the well-established relationships between body size and cranial volume among hominids would be maintained. If LB1 were a dwarfed Homo sapiens with an estimated body mass of 16–27.8 kg, the expected cranial capacity, taking brain-body mass data and the scaling exponent value for a modern European population (Holloway, 1980), would be 1,109–1,223 cc (range, 817–1,604 cc; Table 4). Using the same intraspecific scaling exponent, the expected cranial capacity if LB1 were a dwarfed Homo erectus, taking cranial capacity data from Stanyon et al. (1993), would be 794–876 cc (overall range, 583–1,107 cc; Table 5). Based on the Dmanisi specimens, the expected cranial capacity of a similarly dwarfed individual from this population would be 560–662 cc (overall range, 495–706 cc; Table 5). The cranial capacity of ∼ 400 cc of LB1 would be 5.4–6.2 standard deviations below the expected value for Homo sapiens. Using the same coefficient of variation for brain size, the cranial capacity of LB1 would be 4.2–5.1 standard deviations below the expected value for Homo erectus, and 2.3–3.2 standard deviations below the expected value for the Dmanisi sample (Tables 4 and 5). Among modern humans, severe congenital microcephaly or primary microcephaly is defined as head circumference (a surrogate for brain volume) three or more standard deviations below the mean at birth (Dobyns, 2002) or more than four standard deviations below age and sex means (Woods et al., 2005). The brain volume (∼ 400 cc) and estimated head circumference for the LB1 skull would be more than four standard deviations below the mean for either a dwarfed Homo sapiens or a dwarfed Homo erectus population (Tables 4 and 5).

Table 4. Calculation of brain size of a dwarfed Homo sapiens with the same body weight as LB1
VariableValueNotes on additional calculations
  1. Data and statistics from Holloway (1980) for a large sample of Danish humans (n = 667) carefully selected by removal of cases of pathologies likely to affect brain weight, and extremes of body mass. Some further statistics were estimated or calculated, as indicated.

Modern human data (Holloway, 1980):
Average body weight (g)67100 
Average brain weight (g)1388 
Maximum body weight (g)106000 
Minimum body weight (g)40000 
Maximum brain weight (g)1850 
Minimum brain weight (g)1040 
s.d. brain weight132Calculated from male and female values
Brain scaling exponent0.168 
Brain scaling intercept218 
Brain scaling intercept – using max brain weight286Calculated using brain scaling exponent, maximum brain weight and average body weight
Brain scaling intercept – using min brain weight161Calculated using brain scaling exponent, minimum brain weight and average body weight
Estimated brain weight (g) or cranial capacity (cc) (approximately equivalent (Martin, 1990)) for dwarf Homo sapiens with the same body weight as LB1
At body weight estimate 28.7 kg:
Average brain weight (g) or cranial capacity (cc)1223 
Maximum brain weight (g) or cranial capacity (cc)1604Calculated using brain scaling intercept for maximum brain weight
Minimum brain weight (g) or cranial capacity (cc)902Calculated using brain scaling intercept for minimum brain weight
At body weight estimate 16 kg:
Average brain weight (g) or cranial capacity (cc)1109 
Maximum brain weight (g) or cranial capacity (cc)1454Calculated using brain scaling intercept for maximum brain weight
Minimum brain weight (g) or cranial capacity (cc)817Calculated using brain scaling intercept for minimum brain weight
Difference between the average cranial capacity estimate for dwarf Homo sapiens and the actual cranial capacity of LB1 (400 cc) in standard deviations calculated for brain weight variation for the Homo sapiens sample
At body weight estimate 28.7 kg:
LB1 cranial capacity in standard deviations below expected size for dwarf Homo sapiens6.2 
At body weight estimate 16 kg:
LB1 cranial capacity in standard deviations below expected size for dwarf Homo sapiens5.4 
Table 5. Calculation of brain size of a dwarfed Homo erectus or dwarfed individual from the Dmanisi population with the same body mass as LB1
VariableHomo erectusDmanisiNotes
Stanyon et al.Gabunia et al.; Rightmire et al.; Vekua et al.
  1. Data from Gabunia et al. (2000, 2001), Kappelman (1996), Rightmire et al. (2006), Stanyon et al. (1993), and Vekua et al. (2002).

Average body weight (kg)6050Kappelman (1996); Gabunia et al. (2001)
Average cranial capacity (cc)991664 
Maximum cranial capacity (cc)1251775 
Minimum cranial capacity (cc)727600Calculated using coefficient of variation for brain weight of Homo sapiens sample (see Table 4) and average cranial capacity for Homo erectus sample
Estimated population s.d. cranial capacity9463Using value for Homo sapiens sample
Cranial capacity scaling exponent0.1680.168 
Cranial capacity scaling intercept156.2107.8 
Cranial capacity scaling intercept - using maximum cranial capacity197.3125.9 
Cranial capacity scaling intercept - using minimum cranial capacity114.697.4 
Estimated cranial capacity (cc) for dwarf Homo erectus and Dmanisi population with the same body weight as LB1
At body weight estimate 28.7 kg:
Average cranial capacity (cc)876605 
Maximum cranial capacity (cc)1107706Calculated using cranial capacity scaling intercept for maximum cranial capacity
Minimum cranial capacity (cc)643547Calculated using cranial capacity scaling interceptfor minimum cranial capacity
At body weight estimate 16 kg:   
Average cranial capacity (cc)794548 
Maximum cranial capacity (cc)1103640Calculated using cranial capacity scaling intercept for maximum cranial capacity
Minimum cranial capacity (cc)583495Calculated using cranial capacity scaling intercept for minimum cranial capacity
Difference between the average cranial capacity estimates for dwarf Homo erectus and dwarf Dmanisi individual and the actual cranial capacity of LB1 (400cc) in standard deviations
At body weight estimate 28.7 kg:
LB1 cranial capacity in standard deviations below expected size or dwarf Homo erectus5.13.2 
At body weight estimate 16 kg:   
LB1 cranial capacity in standard deviations below expected size or dwarf Homo erectus4.22.3 

Microcephalic disorders are not particularly rare, as one of us (W.B.D.) has ascertained more than 200 such individuals for study (see also Argue et al., 2006). To date, four human genes have been cloned that result in primary microcephaly with mild to moderate mental handicap and survival well into adult life (Woods et al., 2005). At least two of these (ASPM and MCPH1) have evolved rapidly in hominids and other primates and are hypothesized to have contributed to the rapid increase in brain size shown in Figure 1 (Zhang, 2003; Evans et al., 2004; Kouprina et al., 2004; Wang and Su, 2004; Gilbert et al., 2005). It has been proposed that genes involved in regulating brain size, particularly a subset of microcephaly genes in which mutations produce high-functioning forms of microcephaly, may have undergone advantageous mutations in evolution leading to brain enlargement with few deleterious side effects (Gilbert et al., 2005). Thus, it is certainly conceivable that LB1 could represent a microcephalic individual from a small hominid population.

Alternatively, LB1 could be derived from an extant population of normal hominid stature, more likely Homo sapiens than Homo erectus. Under this hypothesis, LB1 would have a short stature with microcephaly syndrome in which both body size and brain volume are far below the norms for the extant population. Various syndromes with severe intrauterine growth retardation and proportionate (at least at birth) microcephaly have been described in modern humans, including Bangstad, Bloom, Buebel, de Lange, Dubowitz, Kennerknecht, Meier-Gorlin, Okajima, and Seckel syndromes, as well as Majewski (microcephalic) osteodysplastic primordial dwarfism (MOPD) type 1, MOPD type 2, MOPD-Cervenka type, and MOPD-Toriello type (Toriello et al., 1986; Bangstad et al., 1989; Opitz and Holt, 1990; Meinecke et al., 1991; Lin et al., 1995; Buebel et al., 1996; Bongers et al., 2001; Silengo et al., 2001; Faivre et al., 2002; Okajima et al., 2002; Hall et al., 2004). Several of these syndromes are associated with survival to adulthood.

In their original report on LB1, Brown et al. (2004) state without discussion that primordial microcephalic dwarfism in modern humans does not replicate the skeletal features present in LB1. To the contrary, we find such a comparison interesting. The best studied of these syndromes is MOPD type 2. While the reported skeletal features of LB1 differ from this syndrome in several regards, the similarities in overall size are remarkable and instructive. Affected children have severe intrauterine and postnatal growth retardation and microcephaly with normal or mildly impaired intelligence and may survive to adulthood (Hall et al., 2004). They are remarkably small, with weight, length, and head circumference at birth proportionately reduced to the size of a 28-week gestation fetus. Postnatal growth is poor, with head growth much slower even than stature, resulting in an adult height of 100–110 cm and head circumference of 38–41 cm. These values are well in line with the LB1 skeleton. A small jaw with deviant development of the chin and dental anomalies is common, including dysplastic and missing teeth in both primary and secondary dentition. The LB1 fossil also has a small jaw with dysplastic and missing teeth, although, as we have noted, other skeletal changes differ. Development of the chin is highly variable in microcephalics. The chin is particularly prominent in some cases, as in the Stuttgart microcephalic (Figs. 3 and 6), whereas in others the mental eminence is weak or lacking. Dokládal (1958), for example, reported on a 57-year-old microcephalic with a cranial capacity of 405 cc having a small mandible with weak development of the chin.

Seckel syndrome consists of similar intrauterine and postnatal growth retardation and microcephaly, typically more than seven standard deviations below the mean, with moderate to severe mental retardation but frequent survival to adulthood (McKusick et al., 1967; Majewski and Goecke, 1982; Faivre et al., 2002). Skeletal changes are present, but less severe than in MOPD type 2 (Tsuchiya et al., 1981).

While it is not possible to match any of these syndromes exactly with the LB1 fossil based on the limited data available, the features of several are informative. We find that this group of syndromes shares several features with the LB1 fossil, including very similar small stature and head size, a small and receding jaw, and dental anomalies. Lacking the soft tissues and some skeletal components of LB1, we cannot conclude that LB1 had any particular one of these syndromes, but we do think that the substantial overlap in features supports this possibility. One major limitation for comparative studies is the absence of information on the postcranial skeleton in museum specimens of human microcephalics. In closing, it should be noted that the third hypothesis, that LB1 may derive from a more primitive (pre-erectus) population, cannot be addressed by consideration of modern human developmental abnormalities.


We conclude that the features of LB1 best support the interpretation that it is a pathological, microcephalic dwarf specimen of Homo sapiens (see also Jacob et al., 2006). Richards (2006), in a study emphasizing growth processes, reached a similar conclusion that LB1 probably belonged to a modern human population with reduced stature (attributable to a modification in the growth hormone/insulin-like growth factor I axis), but also suffered from a mutation in the MCPH gene family. However, he differs in regarding this combined condition as nonpathological. If further specimens directly resembling the Flores skull with respect to the tiny cranial capacity were to be discovered, the probability that such an explanation is correct might diminish. However, the likely autosomal recessive inheritance of such a syndrome means that such evidence would not necessarily be critical. On the basis of present evidence, it seems most likely that the LB1 specimen is a pathological anomaly, not a new species.

While this account has focused on the LB1 skeleton, because brain size is known only for that individual, some comment is required on the other specimens that have been reported from Flores. These have been interpreted as providing evidence that a small-bodied hominid inhabited Flores at least between 95,000 and 15,000 years ago. As has been explained, the presence of other small-bodied individuals in itself poses no problem. It is the tiny brain size of LB1 that poses a problem. However, the discovery of a second mandible lacking a chin (LB6/1) does raise questions, particularly because it is claimed that it is only 15,000 years old and hence 3,000 years younger than the LB1 skeleton (Morwood et al., 2005a). If the lack of a chin is interpreted as a side effect of microcephaly in LB1, it would be difficult to ascribe this condition in the second mandible to persistence of a rare autosomal recessive condition for 3,000 years on Flores. However, the dating of the second mandible depends on the interpretation that the cave sediments have remained undisturbed and that no intrusive burials occurred. The apparent mingling of at least two different assemblages of stone tools in the deposits suggests that the sediments have not remained completely undisturbed. An alternative possibility is that the LB6/1 mandible is from a small-bodied individual that did not suffer from microcephaly and that the absence of a chin in both known mandibles is in fact a local variant attributable to some other cause. It should be noted that a significantly reduced chin is found in some modern African and Indonesian pygmy populations and Australo-Melanesians (Jacob et al., 2006; Richards, 2006). Furthermore, it should be emphasized that, although the two Flores mandibles are broadly similar in overall size, there are several differences of detail. Unlike that of LB1, the LB6/1 mandible shows no obvious dental anomalies, its dental arcade differs in shape, and the ascending ramus is markedly smaller in height.

Brown et al. (2004) stated that LB1 is megadont relative to both Homo ergaster and Homo sapiens. In fact, examination of the scaling of lower molar teeth area in various hominids compared to a large sample of monkeys and apes reveals that LB1 is similar to typical anthropoids and early Homo, with relatively smaller teeth than the truly megadont australopithecines (Fig. 12). While normal modern humans have relatively small teeth in relation to body size, the mandibular molar area in the Lesotho microcephalic is very close to the value for LB1, at a very similar body mass. Hence, if LB1 is megadont to any degree, so is the Lesotho microcephalic. However, because scaling of teeth follows a similar pattern to the scaling of brain size during dwarfing, individuals with reduced body size would be expected to show somewhat overscaled dental dimensions (Shea and Gomez, 1993). Interestingly, the molars in the LB6/1 mandible are appreciably smaller in area than those of the LB1 mandible, providing a further difference between the two specimens.

Figure 12.

Plot of unilateral summed mandibular molar area for 76 monkeys and apes (nonhominid anthropoids) compared with a sample of hominids. A least-squares regression line has been fitted to the nonhominid anthropoids as a visual guide. As expected, the megadont australopithecines (Australopithecus and Paranthropus) all lie above the line, whereas representatives of Homo lie on the line or below it. Key to points for Homo: 1 = Homo habilis; 2 = Homo erectus; 3 = Tasmanian aboriginal Homo sapiens; 4 = 17th-century European Homo sapiens (London); 5 = Lesotho microcephalic Homo sapiens; 6 = Flores LB1 mandible; 7 = Flores LB6/1 mandible. Molar dimensions for nonhuman anthropoids from Kanazawa and Rosenberger (1989), Lucas et al. (1986), and Swindler (1976); for fossil hominids from Blumenberg and Lloyd (1983); for modern Homo sapiens from Brace (1979); for LB1 from Morwood et al. (2005a). Gary Sawyer kindly provided measurements of the lower molars from the AMNH cast of the Lesotho microcephalic mandible. Body mass values for nonhuman anthropoids from Smith and Jungers (1997) and for fossil hominids from McHenry (1994). Note that the same body mass of 23 kg has been taken for both LB1 and LB6/1.

Argue et al. (2006) recently applied canonical variate analysis (CVA) to compare the skull of LB1 with a comprehensive sample of modern Homo sapiens, two microcephalic H. sapiens, representatives of fossil Homo (specimens attributed to H. erectus and H. ergaster), australopithecines (Australopithecus and Paranthropus), and chimpanzees (Pan paniscus and P. troglodytes). In separate plots of CV1 against CV2 using different data sets, the two microcephalics were found to occupy a peripheral position relative to the general cluster of points for Homo sapiens, while the point for LB1 was distant from that cluster and close to Homo ergaster (notably KNM-ER 3733). In fact, the microcephalics are in both cases located in the general vicinity of LB1, but LB1 is undoubtedly further removed from the general cluster for H. sapiens. As Argue et al. (2006) themselves acknowledge, “microcephaly is an extremely heterogeneous condition and, while our results are suggestive, it may be that they would differ should a larger sample of microcephalics be studied.” In fact, the two microcephalics included in their study are problematic in various respects. Both are archaeological specimens dating back 2,000 y or more, one from Crete (Poulianos, 1975) and one from Japan (Suzuki, 1975), and therefore lack any documentation of their condition. In the Minoan microcephalic skull from Crete, the third molars were not fully erupted, so the individual concerned presumably died before reaching adulthood. For reasons explained above, that skull is therefore not really suitable for comparison with LB1. This objection does not apply to the Japanese skull (Sano 3), which is dentally adult. However, both the Minoan and the Sano skulls have larger cranial capacities than LB1. Unfortunately, Argue et al. (2006) give two different values for the cranial capacity of the Minoan microcephalic: 350 and 530 cc. It is the higher value that is correct. In the case of the Sano skull, the recorded cranial capacity of 730 cc is almost twice the average of about 400 cc for modern human microcephalics and the value of about 400 cc for LB1.

Argue et al. (2006) also considered certain postcranial elements (radius and femur) in comparing LB1 with apes and hominids (though not with microcephalics, as no postcranial elements were discovered with the Minoan or Japanese microcephalic skulls). As had been noted previously, in the LB1 skeleton the arms are unusually long relative to the legs (Morwood et al., 2005a). Taking an unconventional ratio of radius length to femur length, Argue et al. (2006) conclude that LB1 is intermediate between African apes and extant Homo sapiens, being more similar to Australopithecus garhi than to other hominids. It should, however, be noted that the radius is unknown for LB1 and that the length of the radius taken by Argue et al. (2006) was actually inferred from the length of the (incomplete) ulna of LB1 by Morwood et al. (2005a). While it is surely true that the forelimb:hindlimb ratio of LB1 shows some resemblance to a more primitive condition in hominid evolution, the significance of this cannot be properly assessed without information on the condition of the postcranial skeleton in modern human microcephalics. In fact, as noted by Richards (2006), the ratio of forelimb length to hindlimb length (intermembral index) increases with decreasing body size as a consequence of allometric scaling (Shea and Bailey, 1996), although this does not account for the extreme condition found in LB1 (Argue et al., 2006).

It is marginally possible that the hominid remains from Flores provide evidence of a new species from a lineage that diverged at a very early australopithecine stage, about 3 Ma ago, when cranial capacity was still very small. However, this would require convergent evolution of many similarities to Homo species, and the complete lack of documentation of such a lineage in the fossil record represents a major problem. Furthermore, in this case it certainly cannot be argued with any degree of plausibility that Homo floresiensis produced the stone tools found in association with the skeletal remains. On the basis of all the evidence presented here, it seems to us most probable that LB1 was a microcephalic modern human.


Thanks are due to Martyn Cooke for preparation of the endocast of the microcephalic hemiskull at the Hunterian Museum, Royal College of Surgeons, London; Ken Mowbray at the American Museum of Natural History, New York, for providing initial information on the microcephalic skull cast and for measurement of its cranial capacity; Matt Grove at the Field Museum, Chicago, for preparation of an endocast from the original Stuttgart microcephalic skull; Will Pestle for locating the microcephalic skull cast and endocast from Lesotho in the collections of the Field Museum and for measurement of the endocast volume; Jill Seagard for providing drawings for Figures 8 and 9; John Weinstein for producing the photographs for Figure 9; and Michel Hofman for providing access to the data used in his 1984 paper. Dean Falk readily provided information on the AMNH microcephalic skull from which a virtual endocast was generated for her published study (Falk et al., 2005a). We are grateful to Ian Tattersall and Gary Sawyer in the Department of Anthropology, AMNH, for their help in providing access to the microcephalic skull cast and accompanying documentation from the collections in their care. Ian Tattersall kindly provided permission for the removal of minute samples by Gary Sawyer from the cast for chemical analysis. Gary Sawyer also generously prepared an endocast from the AMNH microcephalic skull cast and drew our attention to the key paper by Vogt (1867) that proved to contain a description of the original skull in Stuttgart. Thanks are also due to Jeffrey Schwartz for generating electronic images of the AMNH microcephalic skull cast. Chemical analysis of samples from the calotte and lower part of the cranium of the AMNH microcephalic skull cast was conducted by Laure Dussubieux using an ICP-MS acquired with a grant from the National Science Foundation (PI: P. Ryan Williams). Thanks are due to Laure Dussubieux and P. Ryan Williams of the Field Museum for valuable discussion both in planning of the chemical analysis and in interpretation of the results. Elmar Heizmann kindly provided valuable information, including electronic images, concerning the original microcephalic skull in the collection of the Staatliches Museum für Naturkunde, Stuttgart. Doris Morike in the Zoology Department of the Staatliches Museum für Naturkunde helpfully provided an estimate of the age of that individual and also graciously provided permission for a 6-month loan of the specimen. Edna Davion and Elizabeth Shaeffer provided valuable logistic support at the Field Museum, notably with literature searches, data collection and analysis, and preparation of several figures. Able assistance with final preparation of figures was also provided by Julie Delamare-Deboutteville. Jonathan Brown deserves special thanks for conducting computed tomography of the Stuttgart microcephalic skull and producing virtual reconstructions. Computed tomography of the Stuttgart skull was carried out at Northwestern Memorial Hospital using a Siemens Somaton Sensation 64 CT Scanner, and the support of Northwestern University Feinberg School of Medicine Department of Radiology is gratefully acknowledged. Three-dimensional image reconstruction from CT data was performed with an iView workstation, kindly loaned to the Field Museum by TeraRecon, Inc. (San Mateo, CA). Thanks are also due to Judith Hall for helpful discussions regarding human syndromes characterized by short stature and microcephaly, and to Louise Roth for providing information on dwarf elephants. We are grateful to Robert Eckhardt for providing valuable information and numerous comments, and to Ralph Holloway for sharing his expert knowledge of hominid endocasts.