Peopling island rainforests: global trends from the Early Pleistocene to the Late Holocene

ABSTRACT This paper is a cross-comparative examination of how tropical forested islands were populated by humans. It first describes the unique ecological conditions of these environments, how they fluctuated during glacial cycles, and the challenges and affordances they provided people. The paper then explores the global archaeological record, classifying modes of colonisation that led insular tropical forests to be populated. These modes include terrestrial colonisation followed by insularisation (Mode A), maritime colonisation followed by major landmass reconfiguration (Mode B), maritime colonisation of uninhabited islands that always remained insular (Mode C), and maritime colonisation of already inhabited islands (Mode D). Finally, the paper discusses how, amongst Homo sapiens, ongoing dynamism between human adaptive behaviours and environmental flux stimulated processes of diversification, specialisation, and connectivity in these crucial ecologies; by contrast, archaic hominins like Homo erectus, Homo floresiensis, and Homo luzonensis may have found changes associated with forest expansion and insularity extremely challenging.


Introduction
Islands and tropical forests are two environments with the capacity to isolate and differentiate living organisms, including hominins. In biogeographic terms, insularity constrains the variety of plants and animals present but encourages the emergence of novel endemic species (Whittaker and Fernández-Palacios 2007). On the other hand, tropical forests produce high biomass and biodiversity but low densities of large fauna and carbohydrate-rich plant resources (Corlett and Primack 2011). The combination of these environments -island rainforests -should therefore exacerbate processes of isolation and diversification in peculiar ways. Despite this, there have been few attempts to bring these two environments together on a broad theoretical level to understand the implications for the people that interacted with island rainforests in the deep past. There is scope to explore how hominin adaptive flexibility over the longue durée became interlinked with these novel and perhaps challenging places to live, especially through periods of environmental fluctuation.
This paper expressly explores the peopling of rainforested islands at the global scale. It does so by first synthesising biogeographic and environmental studies to identify the distinctive ecological contingencies of these spaces. It then classifies the modes by which tropical forested islands were have important implications for human habitability (Gaffney 2021b). Around the sea, these effects are often related to increased wind velocity and damage to the forest edge and exposed areas of the canopy. The combination of tall canopies and shallow roots makes the rainforest highly susceptible to adverse winds, especially during tropical monsoons and cyclones (Corlett and Primack 2011, 14). This is compounded by the depletion of ground water and increased evapotranspiration near the edge of tropical forests (Bierregaard Jr et al. 1992). In the case of humans, however, edge effects often positively impact mobility and settlement, which are expedited by open spaces along the coast, nearby rivers, or in clearances within the forest .
Forest fragments also affect the fauna and flora available to humans, with different species feeding on successional and secondary plants on the margins, and increased competition brought about by limited resource patches (Gibson et al. 2013). Fruit and nut bearing trees in the under storey of the forest grow more readily around forest edges (Mountain 1991a). In terms of fauna, forest fragments constrain the movement of most forest-adapted species (Malcolm 1988(Malcolm , 1991 and lead to compressed population sizes (Bierregaard Jr & Lovejoy 1989). Despite overall reductions to diversity and abundance on island rainforests, edge effects also encouraged specific animals to congregate at the margins; all of these factors exposed fauna to the impact of humans hunting (Cullen, Bodmer, and Valladares Pádua 2000 400 500 Figure 1. Edge effects at the margins of rainforest fragments, displaying the distance into the forest fragment each effect is felt (data from Laurence et al. 2002).
Processional forcing and climatic cycling during the Quaternary made island rainforests dynamic, with fluctuations in sea-level, precipitation, and reef development drastically changing their extent and topography throughout archaeological time. As sea levels fluctuated, they connected and separated islands from continental shelves by revealing and then submerging land bridges (Fernández-Palacios et al. 2016). During interglacials, sea level rise submerged large tracts of continental shelf including forested regions. However, increases in atmospheric CO 2 , ambient temperatures, and precipitation simultaneously encouraged the expansion of lowland C3 plants, forming archipelagos of rainforested islands. In Asia, the Palk Strait infilled during the Holocene, cutting Sri Lanka off from India (Banerjee 2000). The Sunda shelf, forming an extension to mainland Southeast Asia, has been repeatedly submerged and exposed during marine transgressions. This turned Sunda into the islands of Borneo, Java, and Sumatra at the start of the Holocene, although the Philippines and Wallacea had always remained insular, separated from the mainland by deepsea trenches (Sathiamurthy and Voris 2006). To the east, the islands of New Guinea, Aru, and parts of the Raja Ampat and Massim groups were attached to Australia during low-stands, comprising the continent of Sahul, with the Gulf of Carpentaria at times forming a large paleo-lake (Torgersen et al. 1983). Off Sahul's northeast coast, the Bismarck Archipelago and Solomon Island chain were always separated from the mainland. Throughout human history, Madagascar has remained an isolated continental island, experiencing small fluctuations in size (Ali and Huber 2010). Similarly, except for Trinidad and Tobago, the Greater and Lesser Antilles of the Caribbean were never connected to the mainland, but many Pleistocene islands were submerged or reduced in size at the start of the Holocene (Cooke et al. 2017).

Patterns in the peopling of rainforested islands
We can identify patterns in the coeval climatic processes and behavioural trajectories that led island rainforests to become populated. Here, I group these into four peopling processes or modes ( Figure 2). Mode A refers to a process of initial terrestrial occupation followed by insularisation, leading pieces of forested land and their human inhabitants to be cut off from the mainland. Mode B refers to large landmasses initially being peopled by water crossings, after which they were insularised. This process is behaviourally distinct from Mode A because it involved seafaring practices emerging prior to insularisation. Mode C refers to water crossings leading uninhabited forested islands to be populated, but despite sea level fluctuations the islands remained insular during the Pleistocene, or the islands were reached after Holocene marine transgression. Mode D refers to water crossings to already-inhabited marine islands, leading to overlapping and distinct cultural groups living on the same island. This classification scheme is now used to critically assess the global archaeological dataset.

Mode A: insularisation
Homo erectus was present around Sunda from about 1.3 million to perhaps c.100,000 years ago (Matsu'ura et al. 2020;Rizal et al. 2020) (Figure 3). During this time, the shelf experienced major fluctuations to its shape and size, at times turning what is now Java into a peninsula and, particularly in Marine Isotope Stage (MIS) 5 (130,000-80,000 years ago), an island (van den Bergh et al. 1996). Based on what we know about H. erectus mobility and niche selection, it is likely that these hominins moved into Java during successive sea level lowstands at c.1.8 million years ago and again c.1.2 million years ago (Sémah et al. 2010). During these times, mosaics of open tropical forests  and savannah were dominant, lagoons formed around volcanic domes, and mangroves were present around coastlines (Bettis et al. 2009;Sémah and Sémah 2012). There is evidence that H. erectus foraged for shellfish (Joordens et al. 2009) and reused the shell remains for tools and mark making (Choi and Driwantoro 2007;Joordens et al. 2015); in these coastal and lacustrine areas H. erectus may have even experimented with rudimentary swimming and rafting (Gaffney 2021a).
There is yet to be direct evidence for H. erectus living in the dense tropical rainforests that expanded during interstadials (Roberts 2019, 88), although forest margins could have plausibly been targeted (Sémah et al. 2016). It may not be a coincidence that the latest evidence for H. erectus on Java in MIS-5 corresponds with a period of rapid fluctuation, insularisation, and high precipitation (Cutler et al. 2003). Around Sunda, this period was associated with the increased temperature and humidity that prompted the expansion of tropical forests (Van Der Kaars and Dam 1997) and faunal turnover leading grassland and open woodland species to be supplanted by rainforest varieties (Storm et al. 2005). H. erectus may have been unwilling and unwitting participants in the final stages of these Mode A processes, encountering difficulties in navigation and subsistence maintenance under the canopy, and suffering gradual population bottlenecks having been cut off from mainland Southeast Asia.
By contrast, the earliest evidence for Homo sapiens on Sunda at Lida Ajer dates to MIS-4, 73,000-63,000 years ago (Westaway et al. 2017). This was a period of lowered sea levels and, most parsimoniously, our species entered the region over land. However, following the LGM, the insularisation of formerly continental savannah and the expansion of lowland rainforests turned Sundalanders into rainforest-islanders. On large newly formed islands, humans took advantage of rainforest expansion rather than totally abandoning habitation sites. Sites like the Niah Caves demonstrate that some parts of northern Borneo were continually surrounded by patches of closed canopy rainforest in the Late Pleistocene (Stimpson 2012), and people hunted a variety of rainforest fauna throughout the period of insularisation, particularly including large ungulates like wild suids, as well as orangutan, gibbons, monkeys, fruit bats, forest birds, rodents, and sun bears (Arifin 2017;Piper and Rabett 2009). The encroachment of the sea to the north, however, did lead to the development of mangrove forests that encouraged increased foraging for large soft-shelled turtles (Barker et al. 2011;Pritchard, Rabett, and Piper 2009), while warming temperatures facilitated hunter-gatherers to frequent the dense rainforested interior of the island as well (Kusmartono, Hindarto, and Herwanto 2017). Similarly, when Sri Lanka separated from the subcontinent, sites like Fa-Hein Lena demonstrate that humans continued to hunt in the lowland rainforests from MIS-3 to MIS-1, where they specialised on small and hard-to-catch game like monkeys and squirrels Wedage et al. 2019). Composite bone tool production, present throughout the Late Pleistocene to Mid Holocene on both Borneo and Sri Lanka, suggest specialised rainforest hunting practices continued unbroken, and actually intensified, despite the insularisation of these large shelf islands (Barton et al. 2009;Perera, Roberts, and Petraglia 2016;Rabett and Piper 2012).

Mode B: overwater movement and landscape reconfiguration
H. sapiens moved from Sunda to Sahul via Wallacea (to which we will return) certainly before 50,000 years ago (O'Connell et al. 2018) and very possibly before 65,000 years ago (Clarkson et al. 2017) (Figure 4). In doing so, watercraft were used to enable deliberate movements of people (Balme 2013), leading many hundreds of individuals to gradually arrive on the shores of what is now submerged shelf off the northwest coast of Australia and/or on the Bird's Head of New Guinea Bradshaw et al. 2019;Norman et al. 2018). These movements were not unidirectional and people would have returned to their home communities in Wallacea to tell stories of their discoveries and recruit individuals to establish viable founder populations. These factors suggest that the initial phase of frequentation was associated with strong kin and marriage ties between Wallacea and Sahul that operated for several centuries or a millennia before being replaced by networks operating primarily within Sahul.
Once on Sahul, populations likely spread quickly and across large distances into a myriad different ecological zones Crabtree et al. 2021). The Madjedbebe humans had already moved hundreds of kilometres inland from the northwest Sahul coast by 65,000 years ago (Clarkson et al. 2017), and there is currently little evidence for humans using arid coastlines again until around 30,000 years ago (O'Connor and Veth 2008). By contrast, in the Ivane Valley of northeast Sahul, hunter-gatherers were moving between lowland and montane rainforests from sometime before 42,000 years ago (Summerhayes et al. 2010), and the lowland and hilly rainforests around coastlines, like at Bobongara and Lachitu, continued to be visited (Groube et al. 1986;O'Connor et al. 2011a). This initial phase of occupation was associated with complex plant processing in the montane and lowland rainforests of what is now New Guinea (Summerhayes et al. 2010), and the open tropical forests and savannah of what is today Australia (Florin et al. 2020(Florin et al. , 2021. Plant foods were probably supplemented with medium to large bodied marsupial hunting, although the evidence is scarce (Field et al. 2008;Mountain 1991b).
Upon insularisation after the LGM, many areas of Sahul were slowly abandoned. The separation of the Aru Islands from northern Sahul was associated with changes from open grasslands to wetter forest conditions (Hope and Aplin 2005), leading to the abandonment of Liang Lemdubu, which had formerly been used as a base camp to hunt a variety of wallaby species (O'Connor et al. 2002). Similar trends can be seen around Australia, where the newly formed Barrow Island (Veth et al. 2017), High Cliffy Island (O'Connor 1994), Kangaroo Island (Lampert 1979), and the Furneaux Group (Sim 1998), came to be depopulated in the Early and Mid Holocene. By contrast, on the Torres Strait Islands (formerly hilltops surrounded by a large plain) archaeological evidence suggests people arrived in the Early Holocene following insularisation, and that they began to specialise in hunting large sea mammals (David et al. 2004;Wright 2011). Moreover, rainforested inland areas like the New Guinea Highlands and the Bird's Head Peninsula continued to be frequented by foragers (Gaffney and Denham 2021). In some of these places, lowland-highland seasonal mobility likely continued until the terminal Pleistocene (Gaffney, Ford, andSummerhayes 2015, Gaffney 2021a). Those groups frequenting southern New Guinea would have been acutely aware of the major changes that were taking place to the shoreline, as the Arafura and Torres plains flooded, creating the island of New Guinea. People visiting the north coast may have been totally unaware that they were becoming islanders, with only small alterations occurring to the steep northern shoreline (until information spread between highland groups).

Mode C: overwater movement to truly insular rainforests
The first hominins to disperse into tropical forested islands proper moved from Sunda to Flores over 1 million years ago (Brumm et al. 2010), to Sulawesi sometime between 780,000 and 200,000 years ago (van den Bergh et al. 2016a), and to the Philippines before 777,000-631,000 years ago (Ingicco et al. 2018). Although these islands were reconfigured in shape and size during the Pleistocene (Figure 5), the changes were minor compared with flooding of the Sunda and Sahul shelves. Insular and forested conditions had important impacts on hominins throughout climatic fluctuations, producing small-bodied and small-brained Homo floresiensis and Homo luzonensis (Brumm et al. 2016;Détroit et al. 2019;van den Bergh et al. 2016b). It is unclear what the 'Denisovan' branch of hominins looked like that may have lived somewhere in western Wallacea (Teixeira et al. 2021). What is clear is that substantial morphological variation existed amongst these hominins in the insular tropics, mimicking what we see in non-hominoid primates (Foley 1991), and likely reflects limited geneflow between mainland and island populations (Leppard 2015).
Despite occupying these insular environments for one million years, either continuously or in a recurring series of colonisation and extinction events, there is little evidence that Early-Middle Pleistocene hominins substantially adapted their behaviour to island life. For instance, there is no evidence that marine resources like fish and shellfish were incorporated into the diet (Gaffney 2021a;O'Connor et al. 2017a;Rabett 2018), although it is important to note that none of the sites currently on record were coastal during Early to Middle Pleistocene occupation. It is possible, however, that hominins were exploring closed canopy rainforests on these islands. The hominins hunted megafauna including stegodon, rhinoceroses, cervids, suids, and varanids, which suggests the islands were mosaics of grassland, open woodland, and tropical forests ( Figure 6). Stegodon likely thrived in dense closed canopy tropical forests (Ma et al. 2019;Zhang et al. 2017), while varanids tend to inhabit more open woodlands, savannah grasslands, and, particularly, dry deciduous rainforest (Imansyah et al. 2008). Several cervid remains associated with H. luzonensis display cutmarks, akin to that produced using bamboo knives (Manalo 2011), which may further suggest explorations of tropical forest. However, the absence of familiar megafauna likely prevented hominin populations establishing themselves in the smaller and faunally depauperate islands east of the Weber biogeographic line, where animals tend to be small, agile, arboreal, and nocturnal and therefore difficult to capture ( Figure 6).
Several discrete H. sapiens populations dispersed into insular tropical rainforests from MIS-4 to MIS-3, first in Wallacea pushing beyond the Weber Line (O'Connor 2010), then in the Philippines      (Fuller et al. 2011) in the Late Holocene. In these contexts, H. sapiens made longer and more frequent water-crossings than previous hominins had, employing what must have been well-developed seafaring and wayfinding practices to move out of sight of land to reach, for example, Manus (Fredericksen, Spriggs, and Ambrose 1993), the Talaud group (Tanudirjo 2005), and Grenada (Wilson 2007, 45).  Table S1). This includes the Ngandong fauna from Java, associated with H. erectus at Trinil and Sangiran, the Mata Menge fauna from Flores associated with an ancestor to H. floresiensis, and the Talepu fauna from Sulawesi and the Kalinga fauna from Luzon, both associated with an as-yet unidentified hominin. Base map shows present coastline transposed over a -120 m shoreline, which can be considered an approximate maximum lowstand throughout the Middle Pleistocene. Initial Mode C movements followed long-term behavioural trajectories. As shown in Table 1, upon arriving to Sulawesi H. sapiens clearly hunted rainforest fauna (Aubert et al. 2019;Brumm et al. 2018;Simons and Bulbeck 2004) that were, in many ways, similar to those encountered in the equatorial forest refugia of Sunda (Piper and Rabett 2016 However, as H. sapiens moved into smaller island rainforests, the earliest faunal assemblages suggest colonists increasingly employed broad-spectrum foraging practices (Piper and Rabett 2014). Owing to the reduced diversity of available game, they intensively targeted a number of rainforest and littoral resources (Chinique de Armas et al. 2019). Interregional commonalities in these resources included fruit bats which could be found in large caves, along with small and arboreal forest game such as monkeys, squirrels, possums, cuscus, and rodents. Some groups in Wallacea may have even developed pelagic fishing technology by MIS-3 (O'Connor et al. 2011b) and incorporated substantial marine resources into their diets (Roberts et al. 2020). Although small forested islands can pose an existential risk to terrestrially adapted groups, those populations with a long-term history of exploiting the littoral and pelagic zones would find small islands ideal, for instance, in the later expansion of Pleistocene groups onto Talaud (Ono et al. 2015), Gebe (Bellwood et al. 1998), andKisar (O'Connor et al. 2019). It is in these areas, particularly in eastern Wallacean islands, that we find the highest marine ecodiversity on the planet, with an abundance of marine and coastal resources.
Part of what facilitated expansions into increasingly small and biotically depauperate islands was people's niche constructive tendencies (Fitzpatrick and Giovas 2021), which made these ecologies more familiar to incomers. For instance, after several millennia of movements around the Bismarck Archipelago, people began to translocate marsupials from mainland Sahul for hunting (Heinsohn 2003). In the small islands of Remote Oceania, human movements to new islands more rapidly led to the import and cultivation of exotic tree species (Huebert and Allen 2020), as well as clearances for agricultural crops and the transport of animals for hunting and husbandry (Anderson 2009).

Mode D: palimpsests of dispersal
During the Mid to Late Holocene, some technologically Neolithic seafaring groups dispersed into islands that had been occupied for thousands of years by foragers, fishers, and agroforesters, adding to the palimpsest of cultures, languages, and technologies. This occurred as Austronesian speaking groups moved into areas populated by Non-Austronesian speakers in Wallacea and Near Oceania during the fifth and fourth millennia before present (Bellwood 2017, 218, 267;Kirch 2017, 74;Matsumura et al. 2018;Oliveira et al. 2022;Skoglund et al. 2016), as mainland African populations interacted with Malagasy on Madagascar during the third and second millennia before present (Ekblom et al. 2016;Pierron et al. 2014), and as pottery making Saladoid populations encountered, and in many places replaced, 'Archaic' Casimiroid and Ortoiroid groups in the Caribbean during the third millennium before present (Fernandes et al. 2020;Keegan 1994;Reid et al. 2018). Similar processes may have taken place as various iron-using South Asian populations moved to Sri Lanka in the third millennium before present and gradually intermarried with ancestral Vedda groups (Kulatilake 2016;Papiha, Mastana, and Jayasekara 1996). In many of these settings, it is clear that dynamic population movements, mixing, and replacements had occurred throughout the Table 1. Key sites with associated faunal remains reflecting initial Mode C occupation. Primarily terrestrial faunal assemblages are marked in green, primarily marine assemblages are marked blue, and mixed assemblages are marked orange. Note, these lists omit the important terrestrial plant food component of the diet.
Pleistocene and Early Holocene, particularly visible in genetic studies and possibly even involving multiple species in Island Southeast Asia (e.g. Larena et al. 2021;Purnomo et al. 2021). Novel material culture associated with Neolithicisation, however, make the Mid to Late Holocene dispersal horizons archaeologically salient. Incoming populations were sometimes accompanied by totally new agricultural systems. For instance, rice farming was introduced from mainland Asia and Taiwan into parts of Island Southeast Asia (Bellwood 2011). In other settings, crops already established in the islands were processed in different ways, as in the case of Saladoid populations introducing new types of ceramics and griddles for cooking cassava and breads (Hofman, Ramos, and Jiménez 2018). On top of this, new domesticated animals were translocated around islands, as in the case of pigs, dogs, and chickens being introduced into the Pacific from Island Southeast Asia (Summerhayes et al. 2019). The gradual domestication of tropical plants and animals, generated both from long-term cultivation practices and wholesale introductions from continental areas, represents a new threshold in how humans interacted with rainforested islands (Roberts et al. 2017a). This often came alongside forest clearances in order to make space for gardens, agricultural field systems, and expanding settlements, even further fracturing island rainforest fragments. These changes are marked in palaeoenvironmental records by rapid increases in biomass burning (Haberle, Hope, and van der Kaars 2001;Kirleis, Pillar, and Behling 2011;Krigbaum 2003;Le Blanc 2017), and floral and faunal extinctions, which took place, in particular, on smaller oceanic islands (Fall 2010;Olson 1989;Prebble and Dowe 2008;Steadman 1995), but also on larger islands like Madagascar (Hixon et al. 2021). What made Mode D peopling processes comparably unique, was that the advent of new migration flows between continental zones to forested islands led communities to mix with, supplant, and even avoid each other; that is, the presence of existing groups and interaction networks actively shaped how incoming groups moved into rainforested islands. In some areas for instance, in the Bismarck Archipelago, incoming Lapita groups were able to settle themselves at arm's length from established islander groups by occupying marginal, uninhabited islands off the coast of New Ireland and New Britain (Specht 2007;Spriggs 1997, 88). Lapita seafarers may have even bypassed the Solomon Islands owing to the presence of existing populations (Sheppard 2019). These seafarers maintained social and genetic separation over several centuries by reorienting their mobility strategies to frequently cross large stretches of ocean and maintain trade and marriage links with related groups (Lipson et al. 2018;Shaw et al. 2022;Skoglund et al. 2016). As a proxy for this, obsidian from the Bismarck Archipelago was widely distributed to Lapita groups as far east as Fiji (Best 1987), and in the opposite direction to contemporary potting communities on Borneo (Chia 2003). Keegan and Diamond (1987) characterised these seafaring groups as 'supertramps': excluded from larger landmasses by established populations but able to make use of a wide range of resources on several smaller islands. This demonstrates how distinct communities of humans can behaviourally differentiate themselves to carve out different niches in the same ecological space.

Human adaptation and the peopling of island rainforests
The diversity and dynamism engendered by island rainforests had important implications for how hominins moved through, lived in, transformed, and adapted to these ecological spaces. During the Pleistocene and Holocene, humans have come to inhabit highly diverse island rainforest environments such as large islands with dense lowland evergreens, rugged interiors covered in montane rainforest, and small peripheral islands with patchy forest fragments (Keegan and Diamond 1987;. In keeping with classical biogeographic models (e.g. MacArthur and Wilson 1967), insular rainforests compounded the adaptive challenges produced by isolation, remoteness, and the unevenness and unpredictability of resources, and eventually generated exceptional novelty in human behaviour, language, and physiology. For many foraging groups, these contingencies encouraged a kind of synchronic dynamism, involving frequent movement under the canopy or around the coast, and, in many H. sapiens populations, interisland connectivity during initial stages of the colonisation process and at times of population growth.
However, the ontogenic processes that shaped human behaviour and these wider ecologies were also highly variable. Expanding upon recent paradigm shifts in the general dynamic theory of biogeography (Ávila et al. 2019;Borregaard, Matthews, and Whittaker 2016;Fernández-Palacios et al. 2016;Whittaker, Triantis, and Ladle 2008), it is clear that rainforested islands experienced repeated cycles of insularisation and forest expansion and contraction, which presented people with unique and often productive affordances around shifting forest perimeters and shorelines. Diachronically, this produced a continuous dynamism between environmental flux and human behavioural adaptation. In many ways, these developmental pathways played out differently amongst regional hominin populations, and made island rainforests distinct from rainforested continental zones.
Amongst H. sapiens, insularisation encouraged the efflorescence of behavioural adaptive shifts, at times including abandonment, but more commonly, at least on larger rainforested islands, involving persistence and intensification of established rainforest foraging or maritime traditions. In several settings, the movement through a range of forest ecotones, particularly from the coast to rainforested interiors, seems to have been key to maintaining an aggregate of reliable and varied resources amongst mobile hunter gatherers. This is in contrast with H. erectus which became extinct when forests expanded and islands took shape. Perhaps this species was not able to shift regularly between a wide array of foraging strategies or devise ways to take advantage of novel subsistence opportunities when they were encountered.
In a similar way, movements into marine islands led H. sapiens initially to improvise and diversify, but then specialise on maritime resources or small native island fauna and flora, pushing the envelope of what constituted habitable spaces for our species. Initial expansions probably involved frequenting islands by boat for exploration and resource acquisition (speculatively, to acquire fruit bats, birds, small docile mammals, marine animals, and endemic plants) prior to sustained colonisation. That is, there are advantages that uninhabited islands could provide mobile foraging expeditions, and incorporating such spaces into growing archipelagic resource networks was probably a key factor in acclimatising humans to the conditions of small and depauperate forested islands. By occupying these islands regularly, and learning to live on them over many generations, it then scaffolded the possibility for future expansions, allowing later peoples to push into tiny coral atolls, small volcanic islands, temperate and even, on occasion, sub-Antarctic waters. On the other hand, Homo floresiensis, Homo luzonensis, and their ancestors lived on tropical forested islands, and possibly made use of rainforest zones, but did not dramatically reconfigure their subsistence strategies beyond those practiced on Sunda, did not push into small islands (<10,000 km 2 ) lacking megafauna, and became extinct at some time in the Late Pleistocene.
As this paper demonstrates, adaptive flexibility in the tropics emerged in tandem with ecological flux both inside and out of Africa. These developing capacities enabled, and were further encouraged by, movements into novel and challenging environments, including island rainforests. Amongst H. sapiens, innovation and improvisation during initial stages of colonisation were often followed by long periods of specialisation, as humans became increasingly attuned to, and enskilled within, these environments. Specialisation, facilitated by tightly regulated learning systems, is sometimes viewed as 'stasis' in the archaeological record, but it can be highly adaptive and went hand in hand with diversification in island rainforests. Lastly, human tendencies to reshape the environment acted as adaptive buffers, bringing these novel environments closer to the familiar, and allowed humans more readily to engage with these new worlds.

Note
1. Throughout the text, I use the terms peopling, populating, and colonisation in the biogeographic sense to mean the ways that humans established themselves in a new environment.