In the 1960s, palaeoanthropological works focusing on the taxonomy and nomenclature of Hominidae echoed the former works of Dobzhansky (1944) and Mayr (1950), like for instance those of Campbell, 1963, Campbell, 1965 and Campbell, 1972. These studies tended to follow the “syncretistic” school of classification developed in the framework of the Modern Synthesis but, with time, palaeoanthropologists abandoned this framework of classification (Hublin, 2014). What is true for classification is also true for phylogeny (Zeitoun, 2015), as expressed by Bonde (2012): “Now this has been known as a methodological fact for nearly fifty years since Hennig's book Phylogenetic Systematics (1966), so how come that some ‘scientists’ today can be completely ignorant of that, and how can that be accepted by reviewers of professional journals? Is it really impossible to convey such simple and logical biological methodology to anthropologists–or don’t they care?” Nowadays, the spectrum of current palaeoanthropology is sometimes between ignorance and rejection of cladistics (see below).

The cladistic (phylogenetic) school of thought began to influence some palaeoanthropologists in the 1970s. At that time, debates about diagnoses founded on the “total morphological pattern” were replaced by discussions about the value and polarity of the observed traits. In this way, some authors were influenced by cladistics, which devises cladograms based exclusively on derived characters. Thus some palaeoanthropologists proposed lists of such derived characters (Hublin, 1978, Hublin, 1986, Stringer, 1984 and Wood, 1985), others only proposed cladograms (Collard and Wood, 2001, Delson et al., 1977, Johanson and White, 1979, Olson, 1978, Santa-Luca, 1980, Skelton and McHenry, 1998, Strait and Wood, 1999, White et al., 1981, Wood, 1981, Wood, 1992b and Wood and Collard, 1999) and finally, both lists and cladograms were proposed together by some authors (Andrews, 1984, Andrews and Martin, 1987, Eldredge and Tattersall, 1975, Groves, 1989, Hoffstetter, 1974, Lieberman et al., 1996, Skelton and McHenry, 1992, Strait, 2001, Strait and Grine, 2004, Strait et al., 1997, Wood, 1984 and Wood, 1992a). But, apart from rare studies (Chamberlain and Wood, 1987 and Stringer, 1987), data matrices were rarely available in the palaeoanthropological literature.

In a following phase, discussions about the lack of unequivocal results due to disagreements about the scoring of operational taxonomic units (OTUs) and characters on one hand, and, on the other hand, debates about the methods dealing with the variability or the redundancy of some traits (Andrews and Martin, 1987, Bräuer and Mbua, 1992 and Skelton and McHenry, 1992) raised objections that subsequently led palaeoanthropologists to abandon cladistics. Habgood (1989) and Trinkaus (1990) contested the use of cladistics in palaeoanthropology and, in spite of answers by Tassy (1996) to these objections, cladistics was largely ignored or misunderstood by palaeoanthropologists (see Hlusko, 2004). This rejection of cladistics in palaeoanthropology was clearly anticipated by Bonde (1981) and finally, few cladistic analyses of hominids providing a matrix were published in the last two decades (Argue et al., 2009, Asfaw et al., 2002, Caparros, 1997, Curnoe, 2001, González-José et al., 2008, Mounier and Caparros, 2015, Prat, 2002, Smith and Grine, 2008, Widianto and Zeitoun, 2003, Zeitoun, 1996, Zeitoun, 2000, Zeitoun, 2001, Zeitoun, 2009 and Zeitoun et al., 2010).

In an attempt to settle the decade-long debate about Homo floresiensis, we present a thoroughly documented cladistic analysis of the LB1 calvaria.

The phylogenetic pattern of human fossil specimens (as of any fossils) should be resolved before discussing the evolutionary processes leading to such a pattern. Yet, phylogenetic hypotheses in palaeoanthropology almost exclusively result from the analysis of group of specimens lumped together according to the weight of tradition. Because the grouping of specimens under taxonomic names is controversial, we decided to treat every specimen as a different OTU, which corresponds to the original definition of OTUs (Sneath and Sokal, 1973 and Sokal and Sneath, 1963).

We consider together both metric and nonmetric morphological characters in our analysis. Several techniques have been proposed to translate metrics in cladistics (Almeida and Bisby, 1985, Archie, 1985, Chapill, 1989, Mickevich and Johnson, 1976, Strait et al., 1997 and Thiele, 1993), but it would take too long to explain each of them, along with their limitations. Apart from Strait et al. (1997), who proposed a solution, the main problem of these techniques is that they are not applicable by considering one OTU for each specimen.

A metric coding sequence was established by comparing the range of values of the metrics for an extant human (Homo sapiens sapiens) sample taken as a standard (Zeitoun, 2000). For convenience we summarize this coding method here. Because we are mainly considering human taxa, we chose an extant human sample to standardize the amount of variation within each state of the characters for each calculated index. From the extreme value (minimum and maximum) gathered for the sample, the amplitude of variation is divided into segments equivalent to the value of the standard deviation (Fig. 1). Each segment therefore corresponds to a state for the character defined by the index. Among all possible indices calculated from 38 classical cranial anatomical landmarks in physical anthropology (Fig. 2), two current anthropological types of indices were considered: type I = (arch AB - chord AB)/chord AB and type II = (chord AB/chord CD). Bearing in mind that the bilateral data are reduced to one single parameter and to bones taken individually, 2264 indices were calculated per complete individual.

Many of the numerous initial possible metric indices (N = 2264) were rejected because no differences were found between apes, australopithecines and Homo (N_rejected = 1919 e.g. 84.7% of the data). Besides, it can still appear that among the morphological characters several have been described for a single structure, such as the cranial keeling, the eyebrow or the temporo-mandibular articulation. From a strictly anatomical point of view, we took into consideration all the anatomical elements that have been discussed by several anthropologists to describe and distinguish different fossil hominid specimens. Our coding of the states of these non-metric morphological characters can also be used for partially preserved specimens.

Argue et al.’s (2009) study took into account only a non-metric morphological dataset and is based on the a priori grouping of specimens that may not reflect their phylogenetic affinities. In that study (Argue et al., 2009), Homo erectus was represented by three specimens (Sangiran 2, Sangiran 17, Trinil 2), Homo ergaster by two specimens (KNM-ER3733 and KNM-ER 3883), Homo rhodesiensis by a single specimen (Kabwe 1), Homo habilis by two specimens (KNM-ER 1813 and OH 24), and finally, Australopithecus africanus was represented by three specimens (Stw 505, Sts 71 and Sts 5). On the basis of these OTUs defined a priori, Argue et al.’s cladistic analysis suggests that LB1 is the sister group of the australopithecines (e.g. Stw 505, Sts 71 and Sts 5) and that the new taxon Homo floresiensis is characterized by four autapomorphies: the presence of an obelionic depression, the absence of a postglenoid process, an orifice of the incisive canal that is on a plane with the second premolar, and a P4 Tomes’root.

We disagree that these four characters are diagnostic of Homo floresiensis. We also observed an obelionic depression (character 28, state 2) on LB1, but this feature is present in the modern population as well as in some African human fossils and does not appear to be an autapomorphy in our analysis. The LB1 tympanal contributes almost exclusively to the rear wall of the mandibular fossa (character 81, state 1), but there is a small but transversally stretched out processus postglenoidalis (character 119, state 1). It is weak (character 120, state 0) and it laterally overtakes the end of the tympanal (character 121, state 1). It is rounded in anterior view (character 122, state 1) and triangular in lateral view (character 123, state 0).

The incisive (nasopalatine) foramen at the second premolars is a part of the normal human variation observed among Australo-Melanesian populations and is illustrated in Milicerowa (1955). We did not consider dental characters, but according to Haile-Selassie (2001), a P4 with Tomes’ root is present in the Miocene Ardipithecus ramidus and in another hominin from the Atapuerca site. Indeed, Carbonell et al. (2008) indicated that the C, P3 and P4 of the ATE9-1 mandible exhibit single Tomes’ roots with a groove along the mesiolingual root surface. Moreover, in Jacob (1973) described the Sangiran 8 mandible with a Tomes root on the P3 and, finally Jacob et al. (2006) claimed that the Tomes’roots on LB1 in P3 and P4 are completely within expectations for extant human populations from the region. These data suggest that the characters interpretated as autapomorphies by Argue et al. (2009) have a broader taxonomic distribution.

According to our analysis, the LB1 specimen belongs to a clade that can be named Homo erectus, which contradicts both main alternatives discussed: a microcephalic modern human hypothesis or a new species.

Our hypothesis that LB1 belongs to Homo erectus appears to be compatible with additional anatomical parts not taken into account in our phylogenetic analysis. On the one hand, according to Brown and Maeda (2009), both of the Liang Bua mandibles share symplesiomorphic characters with Australopithecus and early Homo and on the other hand, even if the postcranial remains of Homo erectus are poorly known (Donlon et al., 2006), recent studies of the foot, hand, and shoulder provide evidence that features found in the postcranial remains of the Liang Bua series are similar to the primitive morphology of earlier hominins (Jungers et al., 2009a, Jungers et al., 2009b, Larson, 2007, Larson et al., 2007, Larson et al., 2009, Orr et al., 2013 and Tocheri et al., 2007). In both cases (mandible and postcranial), from a cladist point of view, primitive characters do not provide evidence of close affinities.

Other elements seem to support our conclusion. Following the results provided by Balzeau et al. (2005), the endocast of LB1 resembles the endocast of a juvenile Homo erectus more than that of an adult Homo erectus, according to van Heteren (2008). Moreover, on the basis of the exocranial anatomy, when taking all principal components into account with the use of Euclidean distances, Liang Bua only clusters with Homo erectus (Argue et al., 2006 and Lyras et al., 2009). Gordon et al. (2008) acknowledged that their metric analyses of LB1, in which they found it to be similar to Homo erectus (and, to a lesser extent, Homo habilis), might be affected by scaling relationships for crania as small as LB1. Thus, finally should it be hypothetized that LB1 is a small (possibly pathological) Homo erectus?

Even under our hypothesis, the small size of LB1 raises questions about the underlying evolutionary process. Holliday and Franciscus (2009) conclude that the small body size of LB1 cannot be explained exclusively by the retention of primitive (Australopithecus-like) traits. So, how to explain such a small size?

The biological significance of size variation has long been debated in the context of the origins and evolution of Homo, thus playing a key role in discussions concerning species recognition, sexual dimorphism, life history, and phenomena such as Allen's and Bergmann's rules, secular patterns in diet and nutrition, local adaptation to variations in climate, resource availability, and selection driven by diseases or environmental and behavioural factors (Plavcan, 2012). With a height of a little more than one meter and a cranial volume comparable to that of a chimpanzee, the well-known mechanism of “insular nanism” initially evoked by Brown et al. (2004) is advanced as the most likely explanation by some authors (Meijer et al., 2010 and van Heteren, 2012) for the Liang Bua series. Body size is used as a criterion for erecting a new species among various taxa (Tintant, 1972), as exemplified by the distinction between Cervus simplicidens and Cervus elaphus by Guadelli (1996).

If we examine the raw data again, we observe significant differences of interpretation of body size of the Liang Bua series among authors (Table 3). Several human fossils have small brains, and sometimes this criterion has been used to erect new taxa. The KNM-ER 1813 specimen was considered the paratype of the species Homo ergaster by Groves (1989), but because of its brain size (510 cm³), Fergusson (1995) erected the name Homo microcranous, based on this specimen. Notwithstanding, nobody followed Fergusson and, in other human fossil series, some specimens have a small cranial volume but authors did not systematically erect new species for these. The human series of Dmanisi was initially attributed to Homo ergaster (Gabunia, 1992), but the discovery of a large mandible (D2600) in this assemblage was used to erect the species Homo georgicus (Gabounia et al., 2002). Subsequently, because of the discovery that a complete small (546 cm³) calvarium (D4500) and a large mandible (D2600) belong to the same specimen, Lordkipanidze et al. (2013) abandoned Homo georgicus to designate the Dmanisi series. Thus, they put Homo georgicus into synonymy with Homo ergaster. We also suggest abandoning the name Homo floresiensis to designate small Homo erectus and we recommend putting Homo floresiensis into synonymy with Homo erectus.

On the basis of the overall shape analysis of the cranial morphology of several fossils, Lyras et al. (2009) also postulate that the Liang Bua series is distinct from modern humans, including microcephalic specimens, Neolithic specimens from Flores Island, and Melanesian specimens. LB1 cannot be assigned to a modern microcephalic human on the basis of its endocranial (Falk et al., 2005a, Falk et al., 2005b, Falk et al., 2007 and Falk et al., 2009) or exocranial (Argue et al., 2006) anatomy. Eckhardt et al. (2014) considered that the new species hypothesis results from the conflation and misinterpretation. Henneberg et al. (2014) suggested that LB1 had a disturbed developmental homeostasis reflecting Down syndrome. Our phylogenetic results suggest a third explanation.

It is a cardinal rule of systematics that a new species can be erected only from a non-pathological specimen, and this question has to be considered for LB1. Even determining the stature of the Liang Bua series is challenging; Eckhardt et al. (2014), for instance, considered that it is a mistake estimate the stature of LB1 by extrapolation, using the femur length with a formula derived from an African pygmy population. The most astonishing feature of the anatomy of the Flores specimens is not its height (see Jacob et al., 2006 and Henneberg et al., 2014), but rather, its tiny cranial capacity (Brown et al., 2004, Conroy and Smith, 2007, Falk et al., 2005a, Falk et al., 2005b, Falk et al., 2007, Kubo et al., 2013, Martin et al., 2006a, Martin et al., 2006b and Richards, 2006).

For Homo erectus sensu lato, height has been estimated at around 1.50 m, with a mean cranial capacity of 904 cm³ (Rightmire, 2004). Local (Indonesian) Homo sapiens sapiens have a mean height of 1.63 m (Bargain and Zeidan, 2014) and a mean cranial capacity value of 1350 cm³ (Beals et al., 1984). The height of the pathological cases described by Maheshwari et al. (1988) and Murray et al. (2000) among present populations (1.20 m) is comparable to the stature of the Liang Bua series, but not the cranial capacity (1000 cm³, vs about 400 for LB1). Rahmawati et al. (2007) showed that at 5.5 SD below the combined sex mean of the people living at Rampasasa around the Liang Bua cave (1.49 m for males and 1.44 m for females), LB1 is, a clear outlier. Data about the skull size of the entire human data set leads to similar conclusions (Plavcan, 2012). However, the cranial volume of LB1 can be estimated only with moderate precision (Table 3).

Initially, Falk et al., 2005a and Falk et al., 2005b argued that LB1 cannot be considered a microcephal because it is different from their studied specimens, while other authors (Martin et al., 2006a, Martin et al., 2006b and Weber et al., 2005) consider that the former studies could not conslusively reject this hypothesis. From the beginning of this debate, pathologists have suggested various possible pathologies to explain LB1: microcephaly (Martin et al., 2006a and Martin et al., 2006b), Laron Syndrome (Hershkovitz et al., 2007), cretinism (Obendorf et al., 2008 and Oxnard et al., 2010), or microcephalic osteodysplastic primordial dwarfism type II (MOPD II) (Hall et al., 2004 and Rauch et al., 2008).

Brown (2012) asserted that LB1 matched none of the known pathological traits and metrics, and van Heteren (2013) provides convincing and detailed elements to reject the hypotheses interpreting LB1 as a pathological modern human. Nevertheless, Henneberg et al. (2014) gave fresh impetus to the debate by associating facial asymmetry, small endocranial volume, brachycephaly, disproportionately short femora, flat feet and other characteristicistics to Down syndrome. The hypothesis of hypophysal nanism described by Murray et al. (2000), which is a very rare syndrome that leads to an allometrically conservative reduction of the whole individual, has not yet been tested. This pathology is known in Asia and is suspected to occur in the Pacific (Zeitoun et al., 2007). According to Maheshwari et al. (1988), this rare disease occurs at higher frequencies in endogamous populations, and endogamy is favoured by geographic isolation, as on islands such as Flores.