The earliest dispersal of Eurasian hominins is a matter of considerable debate in the literature, not least because relatively few fossil occurrences are available although archaeological sites add to the potential patterning. So far as fossil specimens are concerned, a Later Pliocene/Earliest Pleistocene origin in Africa is perhaps the most widely accepted model for the evolution of Homo (Dennell, 2009, O’Regan et al., 2010, Turner and O’Regan, 2007a and Turner and O’Regan, 2007b), with the earliest known hominin material outside Africa coming from Dmanisi in Georgia (∼1.7 Ma). Specimens from there have been assigned to either a very primitive form of Homo erectus (Rightmire et al., 2006) or a new species H. georgicus (de Lumley et al., 2006), with the latter authors placing H. georgicus closer to the Homo habilis-Homo rudolfensis lineage than to H. ergaster/H. erectus and thus, implying a first Late Pliocene/Early Pleistocene dispersal of most primitive Homo from Africa to Dmanisi.

In Southeast Asia, a date of 1.8 Ma was claimed for H. erectus in Java (Swisher et al., 1994), but material from the top of the Sangiran Formation at close to 1.6 Ma is now considered to be the earliest (Bettis et al., 2009). Homo sp. has been identified at Yuanmou in southern China with a date of 1.7 Ma (Zhu et al., 2008), but this claim has been strongly questioned, most recently by Dennell (2009). Isolated teeth from ‘Ubeidiya in Israel, dated to ∼1.5 Ma (Belmaker et al., 2002), have been attributed to H. erectus or H. ergaster and the record has recently been extended in Europe with the publication of a hominin mandible from the Sima del Elefante at Atapuerca, dated to between 1.2–1.1 Ma (Carbonell et al., 2008).

Earliest Eurasian archaeological remains, regarded as Oldowan (or Pre-Oldowan), also come from Dmanisi (de Lumley et al., 2005). Early Pleistocene sites outside Africa with Oldowan (Mode 1) technology are also found in Spain (Barranco-Leon 5 and Fuente-Nueva 3 and lowest levels of Atapuerca (Carbonell et al., 2008, Gibert et al., 1998, Martínez-Navarro et al., 1997 and Oms et al., 2000), and are claimed at Pirro Nord in Italy at 1.6–1.3 Ma (Arzarello et al., 2007), the Pabbi Hills in Pakistan ∼1.9 Ma (Dennell, 2004), Yuanmou in southern China ∼1.7 Ma (Zhu et al., 2008) and Majuangou in north-eastern China dated to 1.66 Ma (Zhu et al., 2004). We shall return to the question of the success of human occupation in northern China in a later section.

Such a technology, although simple, provides a range of sharp flakes and thus gives a significantly increased range of options for obtaining and processing both foods and other, non-lithic items of everyday use, and the vast numbers of such items found on sites testify to the extent of use and amount of time and effort that must have been put into locating suitable stone. However, the stone tools at ‘Ubeidiya are Acheulean, a form that first appears ∼1.7 Ma at Konso Gardula in Ethiopia (Asfaw et al., 1992), so that its presence at ‘Ubeidiya suggests that at least one hominin dispersal out of Africa occurred between 1.7 Ma and 1.5 Ma and that the populations concerned were equipped with an even more sophisticated technology. Directions and timing of early hominin dispersals into and around eastern Asia have been considered by Antón (2002) and Keates (2004), and clearly, there was no technological impediment to dispersal, although finding suitable sources of stone may have dictated the success of various movements.

Suggesting that Sundaland may have acted as refugium during the LGP, and by implication during previous episodes of lowered sea level, raises the obvious question of a refugium from what? What were the general conditions in Asia during glacial periods? Dennell (2009) has usefully summarised much of the evidence for changes in the climate of the region as a whole during the Pleistocene, and laid stress on the changes in the cyclical fluctuations of the glacial-interglacial cycle that occurred with the MPT, and in particular the increased length and intensity of cold and arid periods and the growth of deserts. Indeed, the increase in these conditions has led him to term the area from north-eastern Africa across Asia and into North China during the latter part of the Middle Pleistocene through until the Last Interglacial “Aridistan” (Dennell, 2009: Fig. 7.20), in contrast with an earlier “Savannahstan” (next section). This shift in conditions would suggest that during the cold periods of the Middle and Late Pleistocene much of the area of central and western Asia was a rather inhospitable place for hominins, while the steppe-like conditions of China north of the barrier produced by the Qinling Mountains (Ferguson, 1993) would have been equally unattractive. This latter point is reinforced by the results reported for the Nihewan Basin by Ao et al. (2010) and Pei et al. (2009) to which we shall return in the next section. We suggest that only Southeast Asia would have provided an area of suitable conditions for the maintenance of long-term populations in Asia during the Middle Pleistocene. Clearly the increase in the size of that area afforded by the sea-level falls in Sundaland adds considerably to the extent of the likely refugium.

But what of conditions during the colder periods prior to the MPT in “Savannahstan”? Dennell (2009) argues that aridity occurred on a much smaller scale at this earlier time, with the differences between glacial and interglacial climates much more muted. In consequence, a larger area of Asia would have remained hospitable during the colder periods, while the area of increased land in Sundaland exposed by the lower amplitude falls in sea levels would have been much less. However, we should also bear in mind that technological developments and abilities of hominins were presumably less the further back in time we go, so that even less extreme conditions might have been rather taxing. There is also the point, well made by Dennell in his concluding remarks (Dennell, 2009: p. 476), that hominin settlement of Asia is always likely to have been constrained and rendered patchy by the need for what he terms the “fortunate coincidence” of stone, water and food prior to the development of long-distance exchange networks that only become obvious in the later stages of prehistory.

Against this background can we offer any general inferences – or hazard a guess – about likely habitats preferred by early hominins, something that we could use as a baseline to assess how attractive or otherwise an area may have been? Dennell (2009) (Table 2.2) has pointed out that hominin localities in Africa prior to 2.5 Ma might be taken to indicate a preference for woodland, although this clearly reflects the history of discovery as much as anything else and finds in Chad (Brunet et al., 1996 and Zazzo et al., 2000) and Bouri (Asfaw et al., 1999 and de Heinzelin et al., 1999) now show an ability to survive in open savannah grasslands.

After 2.5 Ma, it is broadly clear that the development of the global cooling trend that led to the glacial-interglacial cycles produced essentially more arid and seasonal conditions, exemplified in Africa but also seen across continents (Fernández and Vrba, 2006, Turner, 1995, Turner, 1999, Vrba, 1995 and Vrba, 1999). Dennell (2009) (pp. 33–4) has pointed out that the development of more open landscapes has been generally seen as a factor behind the appearance of H. erectus in eastern Africa as a species able to exploit grasslands, and although he mentions a number of caveats to that interpretation Dennell and Roebroeks (2005) had already suggested the term “Savannahstan” as a way of stressing the significant geographic spread of such Late Pliocene and Early Pleistocene habitats from North Africa across into central Asia. However it is worth stressing that savannahs are really tropical grasslands and the term should not be confused with that of grasslands. Savannahs can be defined as landscapes with a continuous grass layer and some trees providing an unclosed canopy cover, with strong alternating wet and dry seasons in tropical and sub-tropical systems (Scholes and Archer, 1997). Outside of the warmth of the tropics the temperate grasslands may mimic features of savannahs, with grazing ungulates, but the grasses are shorter and the variety of trees – especially drought-resistant forms – reduced (Cox and Moore, 1993); indeed the map of Asian grasslands at 3.0 Ma reproduced by Dennell (2009) (Fig. 3.25) clearly shows an area of “savannah, grassland, steppe”. Thus, although earliest hominin dispersals outside of Africa can hardly have avoided crossing and therefore surviving in grassland areas, a preference for savannahs does not automatically translate into a preference for all grasslands, especially in terms of ability to sustain a population in the face of harsher conditions outside the tropics.

The definition of grassland versus the more restricted term savannah lies at the heart of the argument in the paper by Ao et al. (2010), who suggest that palaeoenvironmental evidence from the Nihewan Basin of northern China indicates intensified aridification and cooling leading to reduced tree cover and an increase in what they term open, savannah grasslands. They go on to suggest that the archaeological record from the area can thus be taken to indicate Early Pleistocene dispersal of savannah-adapted hominins into that region. However, the grasslands developed there would seem better characterised as temperate grasslands or indeed steppe rather than savannahs, and it is not clear to us that the archaeological and hominin fossil record from the region supports the inference of a sustained and flourishing population there as some authors propose. We note that the fauna from the Nihewan basin also suggests a grassland or steppe environment (Keates, 2010). Although they avoid use of the term savannah, we argue that the same point can be made in response to the interpretation offered by Pei et al. (2009) for hominin habitats at Donggutuo in the Nihewan Basin during the Early Pleistocene. The latter authors do not argue for continuous occupation, but do conclude that conditions were more benign than we would interpret them to be, with reconstructed average annual temperatures often below, and rarely above, the 7.5 °C of the present day. By way of comparison, we may note that average present-day annual temperatures for the United Kingdom range between 7.0 and 11.0 (Metoffice, 2010), with the highest being in England and in coastal areas, and the averages conceal some often savage winter temperatures that have to be survivable if continuous occupation is to be achieved. Unfortunately, the figures for annual temperature in the Nihewan Basin include no indications of associated errors or temperature ranges, but at the very least it seems to us questionable whether hominins at that time could sustain long-term populations in northern China in such circumstances, since nothing in any savannah-adapted background would seem to have equipped them to survive the cold conditions that would be found in such regions during the colder parts of the year. One might characterise occupation of the Nihewan Basin as a steppe too far for long-term success.

What we have attempted to show here is the potential of Southeast Asia, and in particular the increased land area represented by Sundaland, to act as a refugium for savannah-adapted species, and particularly hominins, during the periods of lowered sea level and environmental changes that characterised so much of the Pleistocene. As Bird et al. (2005) stressed, while precise details of vegetation cover are difficult to reconstruct in the area now covered by seas the presence of open vegetation on sandy areas of exposed land seems very likely, based on the differing lines of evidence that can be brought to bear for the LGM. Louys and Meijaard (2010) have extended this argument to suggest that habitats with open woodlands and grasslands were present during much of the Pleistocene.

For hominins occupying such areas, elements of the fauna such as cervids, bovids and probably suids would seem to have offered an attractive range of resources, as discussed in some detail by Louys et al. (2007). The detailed archaeological evidence for subsistence activities – and the lithic and possible non-lithic tool kit involved – we leave for others to decipher and interpret, but it seems clear that the known range of Southeast Asian stone tools alone would have more than equipped them to deal with medium-sized ungulates, smaller animals and the likely (although unknown) range of other non-animal resources available to them (Dennell, 2009). We therefore believe that the area of Sundaland can be added to the list of likely Asian refugia summarised by Dennell (2009) (pp. 475–6), although further work will be needed to decide whether sufficient quantities of stone were locally available to satisfy the trio of co-incident requirements suggested by him. In particular, we stress the favourable conditions for hominins in Southeast Asia during Middle Pleistocene glacials, conditions otherwise largely absent from much of Asia during this period.

Although the current consensus view is that northern China experienced a continuous level of occupation during the Early and Middle Pleistocene (Ciochon and Bettis, 2009 and Keates, 2010), the result of supposedly benign conditions such as those inferred by Pei et al. (2009) and Ao et al. (2010), we propose that a different scenario may be more plausible for this region. Following on from the “source and sink” population model for Middle Pleistocene Europe put forward by Dennell et al. (2010), we agree that such a model may indeed have the parallels in eastern and southeastern Asia that they suggested. In particular, during the extreme environmental conditions present during the Middle Pleistocene glacials, we propose that northern China may have acted as a population “sink”, in much the same way as northern Europe did. Certainly southern, more warm-adapted mammals, such as Hystrix, Palaeoloxodon and Bubalus, show repeated incursions and excursions into and out of northern China during the Pleistocene, as they did in northern Europe, most likely in response to shifting climatic conditions (Tong, 2006 and Tong, 2007). In much the same way, we propose local hominin population extinctions or contractions would have been counterbalanced by northern re-incursions from Southeast Asia, where environmental conditions, even during the LGM continued to be favourable for hominin populations to flourish.