Aquatic environments were important in the evolution of Homo where abundant shellfish and fish resources were available to provide essential protein and fatty acid nutrients to support hominin brain development (Mellars et al., 2013; Archer et al., 2014; Joordens et al., 2014; Marean, 2014). Archaic species of Homo were clearly able to exploit sessile and easy to gather shellfish from riverine environments since at least 600 kya in Southeast Asia, as seen at Trinil, Java, Indonesia by H. erectus (Joordens et al., 2015). However, efficient exploitation of fish requires complex technologies and sophisticated fishing strategies designed in relation to the life habits of much more mobile and reactive prey than sessile invertebrates (Kirch and Dye, 1979; Masse, 1986; Butler, 1994; Ono and Addison, 2009; Boulanger et al., 2019; Ono et al., 2019; Boulanger, 2021).

The oldest evidence of fishing and fish consumption could have taken place in freshwater habitats of Africa, during the Lower Pleistocene, 1.95 Mya, in the Turkana Basin (Archer et al., 2014). At the site of Olduvai, Tanzania, early hominins may have regularly captured some freshwater catfish species that are easy to catch using very little technology (Stewart, 1994). In the Middle East, a large assemblage dominated by two species of large Cyprinidae (carp) showing evidence of cooking by 780,000 years ago has been recovered at the site of Gesher Benot Ya'aqov, Israel (Zohar et al., 2022). In Europe, it is possible that H. neanderthalensis occasionally exploited freshwater ecosystems and the fish and shellfish species that inhabit them. However, the origins of fish bone accumulations in cave site assemblages associated with this hominin species cannot be unequivocally demonstrated since birds of prey and carnivorous mammals significantly contributed to the accumulation of bones (Kettle et al., 2008; Russ, 2010; Guillaud et al., 2018, 2021). Only the Iberian Neandertals of Figueira Brava seem to have regularly exploited marine aquatic food between 106 and 86 kya (Zilhão et al., 2020). In summary, there is therefore no firm evidence for Neanderthals freshwater fishing (Guillaud et al., 2021).

The advanced cognitive capabilities of H. sapiens significantly facilitated their adaptability to a wider range of aquatic resources and environments before and after their dispersal out of Africa during the late Middle Pleistocene and early Late Pleistocene (Groucutt et al., 2015; Roberts and Stewart, 2018). The earliest unequivocal evidence for hominin fishing is in South Africa ca. 70,000 BP by H. sapiens, both in marine and freshwater habitats (Brooks et al., 1995; Feathers and Migliorini, 2001; Plug, 2006; van Niekerk, 2011; Mitchell, 2015). Yet it is in Island Southeast Asia (ISEA) that the origins of dedicated maritime foraging in the Indo-Pacific probably lie (Kirch and Dye, 1979). The current earliest discoveries, but disputed (see Wood, 2017 for Madjedbebe), of H. sapiens remains in Sunda by 73 kya (Westaway et al., 2017) and in Sahul by 65 kya (Clarkson et al., 2017), indicate the colonization of the insular Wallacean archipelago, situated between these two great continental shelves, was earlier than the oldest known H. sapiens sites (ca. 45 kya) thus far found in the region (Hawkins et al., 2017; Shipton et al., 2019). Zooarchaeological and stable isotope data suggest that early H. sapiens in Wallacea adapted their foraging behavior to marine shellfish and fish in extensive coral reef environments significantly earlier than it was previously thought (O'Connor et al., 2011; Kealy et al., 2020; Roberts et al., 2020, 2022; Boulanger, 2021). Central to this adaptation was a sophisticated and diverse array of marine fishing practices, such as trolling, line fishing, netting or gathering by the use of traps, following lunar cycles and seasons and using a variety of fishing gears such as bone fishing gorges, shell hooks, net sinkers and probably baskets and nets made of vegetal fibers, highlighting the ecological plasticity of our species (O'Connor et al., 2011, 2019; Samper Carro et al., 2016; Boulanger et al., 2019, 2023; Kealy et al., 2020; Boulanger, 2021).

The tropical forests of ISEA were once considered hostile and relatively depauperate in fauna by comparison with coastal habitats (Roberts and Stewart, 2018). However, the presence of early H. sapiens in inland Sumatra, Borneo, and Sulawesi suggests an early adaptation of behavioral and cultural strategies to best meet the challenges of the tropical rainforest environment (Glover, 1981; Barker et al., 2007; Amano et al., 2016; Westaway et al., 2017; Brumm et al., 2018; Ingicco et al., 2020; Ono et al., 2020). There is evidence for exploitation of freshwater shellfish at inland Sunda and ISEA sites as H. sapiens moved into riverine environments on Borneo (Piper and Rabett, 2009), Sulawesi (Glover, 1981; Brumm et al., 2018), and Timor (Glover, 1986; Hawkins et al., 2017) during the early settlement of this region in the Pleistocene. Isotopic evidence indicates an increasing reliance by H. sapiens on a more diverse array of terrestrial resources in Wallacea by the terminal Pleistocene (Roberts et al., 2020).

Freshwater ecosystems may have been crucial to the sustenance and development of foraging cultures moving inland along the waterways on the larger insular islands such as Timor and Sulawesi in Wallacea, as they were in Sunda and Sahul (Barker et al., 2007; Habgood and Franklin, 2008). Freshwater resources were likely more important at inland sites in Wallacea, where terrestrial fauna were relatively depauperate compared to the continental landmasses of Sunda and Sahul (O'Connor and Aplin, 2007). In Sahul, although the anthropogenic origin of sun perch (Macquaria ambigua) otolith accumulations has not been indubitably demonstrated, it is very likely that H. sapiens at Lake Mungo, Australia, consumed them as soon as they arrived on the site 46 kya (Bowler et al., 2003; Long et al., 2018). Early fishing adaptations to inland freshwater habitats at Lake Mungo, speaks to the full modernity of humans throughout their migration from Sunda to Sahul (Langley et al., 2011).

Globally, it appears that fishing using a variety of nets, spears, traps, and lines in these terrestrial aquatic habitats became more widespread during the Pleistocene-Holocene transition (Gorman, 1971; Gobalet and Fenenga, 1993; Piper and Rabett, 2009; Pargeter et al., 2017; Boulanger et al., 2019; Álvarez-Fernández et al., 2020). This coincided with H. sapiens settlement of riverine valleys to engage in sedentary pastoralism, where fishing appears as a secondary activity (Hongo, 1989; Boschian, 1993; Losey et al., 2012; Weber et al., 2013). Eventually carp aquaculture (Cyprinidae) developed around 8,000 years ago in China (Nakajima et al., 2019), but there is no evidence of this spreading to ISEA during prehistory. Only a few sites in the remote islands of the Pacific witness the systemization of inland fishing during the late Holocene (Butler, 2001; Bouffandeau et al., 2019). For a variety of reasons that likely include identification issues, there is little evidence for freshwater fishing practices in ISEA during the Pleistocene (Piper and Rabett, 2009; Hawkins et al., 2017), despite the availability of several endemic edible freshwater fish taxa in wetlands (Zakaria-Ismail, 1994; Göltenboth et al., 2006).

In this paper we investigate H. sapiens fishing practices in Wallacea after inland settlement beside riverine and lakeside terrestrial habitats. To this end we review the scarce evidence for freshwater fish remains from archaeological sites in tropical ISEA while incorporating new evidence from ichthyofaunal bone assemblages from two inland sites in Timor-Leste, Matja Kuru 2 (MK2) and Laili Cave. Laili (4.7 km inland), where human occupation dates between 44.6 and 11.7 kya (MIS 3 and Last Glacial Maximum (LGM)/early Holocene), is situated overlooking a large valley through which the Laleia River runs (Hawkins et al., 2017). MK2 (8 km inland), is dated from ca. 40 kya until the late Holocene but appears to be unoccupied from the onset of the LGM, 28-15 kya (Lambeck and Chappell, 2001) until the beginning of the Holocene. It is situated a few hundred meters from the edge of Lake Ira Lalaro at its full position, the largest freshwater lake in the island of Timor (O'Connor et al., 2014; Roberts et al., 2020) (Figures 1, 2).

Near-shore reef fishing was practiced on the Timor coast both at MK2 and Laili, as shown by the presence of marine taxa such as Belonidae (needlefishes) (Figure 4), Acanthuridae (surgeonfishes, tangs, unicornfishes), Carangidae (jacks, pompanos, jack mackerels, runners, scads), Scaridae (parrotfishes), Scombridae (mackerels, tunas, bonitos), Serranidae (sea basses, groupers) and Balistidae (triggerfishes) (Figure 4; Table 1) presumably incorporating a range of methods that would have perhaps included spearing, netting, and angling to capture fish with different feeding behaviors and habitats (Butler, 1994), thus requiring a good knowledge of local marine environments and appropriate fishing techniques.

Freshwater catadromous and sometimes brackish water taxa, probably adults based on the average size of the bones, are represented at both sites. Gerreidae (mojarras) and Eleotridae (sleeper gobies) dominate the MK2 assemblage. Currently, three species belonging to the genus Gerres (G. filamentosus, G. oblongus and G. oyena) have been recorded on the Timorese coasts and rivers. However, because of the lack of reference material, we were unable to firmly identify the specimens to species or genus level. Most of mojarras are marine and occur in sandy or muddy bottoms bordered by mangroves. They occasionnaly enter rivers and lakes (Froese and Pauly, 2022). Eleotridae could be one of four species (Belobranchus belobranchus, Bunuka gyrinoides, Eleotris fusca and Giuris margaritacea) that have been reported from the south coast of Timor-Leste (De Beaufort and Weber, 1912; Larson et al., 2007). Eleotridae also occur in the Irasiquero River system near MK2, on which Lake Ira Lalaro is located (Figure 1) (Larson et al., 2007). The specimens in MK2 cannot be further identified because of the absence of those taxa in reference collection, added to the fact that we only found postcranial remains. The ecology of Eleotridae is a little more complicated as fish from this family can be found in a variety of environments with a preference for brackish and freshwater habitats, depending on the species and the lifestage of the fish (Bouffandeau et al., 2019; Froese and Pauly, 2022). MK2 and Laili are in close proximity to a freshwater lake and a river where these fish would be readily available and probably caught, as we think it is unlikely that the site's inhabitants would have walked to the distant coast to fish them, albeit we cannot exclude this possibility. A few modern bone fragments have been identified as Cichlidae in the upper more recent spits of MK2. These are consistent with the Mozambique tilapia, Oreochromis mossambicus (Cichlidae), which is endemic to Southeastern Africa. This species was introduced throughout ISEA and Timor-Leste in the 1940's to sustain food production for Timorese populations (Trewavas, 1983). Indeed, the introduction of carnivorous exotic African species such as Clarias gariepinus (Clariidae, north African catfish) and opportunistic omnivores such as O. mossambicus, perhaps modified the Timorese freshwater ecosystems causing endemic species demise (Larson et al., 2007).

Eels, the dominant freshwater taxa at Laili, are present throughout the sequence and four species have been currently recorded in Timor-Leste (Froese and Pauly, 2022). While acknowledging that some taxa in the family Anguillidae may be found in a variety of environments including marine (Froese and Pauly, 2022), we have characterized the Anguillidae in Laili as freshwater, as Anguilla sp. is the only genus that occurs in Timor (A. bicolor, A. celebesensis, A. marmorata, and A. reinhardtii) which only goes to the sea for spawning; this migration pattern remaining locally largely unknown (Arai, 2020). In addition, mullets (Mugilidae), represented by six genera and eight species (Cestraeus goldiei, Crenimugil buchanani, C. seheli, Ellochelon vaigiensis, Osteomugil cunnesius, Planiliza ordensis, P. subviridis, Plicomugil labiosus) today in Timor-Leste have been identified predominantly in the earliest phase of occupation of the site. Mullets can be found in a variety of habitats including coastal marine environments, to brackish and freshwater ecosystems (Froese and Pauly, 2022). Drums or croakers (Sciaenidae), probably of the genus Johnius (as it is the only genus today represented in Timor-Leste with two representatives, J. amblycephalus and J. borneensis) can also sometimes be found in brackish and estuarine environments (Froese and Pauly, 2022).

Today local villagers fish in Lake Ira Lalaro using long flexible bamboo rods equipped with metal hooks. Additionally, invasive species such as tilapia have been introduced to the inland waterways of Timor-Leste. These changes make it difficult to reconstruct traditional fishing techniques. The eels exploited at Laili were probably caught in the adjacent Laleia River using eel pots or nets (Froese and Pauly, 2022). De Almeida (1963) mentions that eels are usually found in springs and streams and, during his anthropological work, recorded that some of them were regarded as sacred by the locals. Sleepers, at MK2 and Laili, could have been fished using traps, weirs or nets (Bouffandeau et al., 2019) made of vegetal fibers in the Irasiquero River system (Larson et al., 2007) or in the Laleia River, while mojarras are usually caught using nets and sometimes line fishing in freshwater/brackish/marine environments (Carpenter and Niem, 1998). Drums or croakers were perhaps caught near the Laleia river mouth by angling or netting.

Different wetland habitats and distance from marine environments have therefore played some role in differences in ichthyofaunal assemblages between MK2 and Laili, with a higher diversity of fished environments at MK2 which is further from the coast and closer to Lake Ira Lalaro (Figures 1, 2). Previously, the most inland prehistoric site on Timor was found at the cave site Uai Bobo 2, located some 20 km from the coast when first occupied ca. 17 kya, but no fish bones were identified there (Kealy et al., 2016) (Figure 1). The new ichthyofaunal assemblages from the inland Timor-Leste sites, Laili and MK2, include small quantities of both marine, brackish and freshwater fish taxa. This indicates exploitation by H. sapiens of a variety of aquatic habitats and some degree of mobility between the coast and hinterland requiring the use of a variety of subsistence strategies culturally adapted to the different environments, as well as planning in terms of resources management. MK2 and Laili are not directly coastal; a return trip from the shelters to the coast on foot can today be made within a day. The Laili fish assemblage was much larger (NISP = 491, MNI = 14) than that of MK2 (NISP = 208, MNI = 38) (Table 1), despite the comparable size of the excavation areas at both sites. This discrepancy may be due to Laili's greater proximity to the coast and river. Although the MK2 fish assemblage is smaller than in Laili, the MNI is higher due to the larger number of Gerreidae pharyngeal plates identified (Figure 4).

Fishing persisted over time at MK2 and Laili, attesting to a detailed knowledge of the aquatic environments passed on over generations through social learning. Nonetheless, as might be expected, fishing at the inland Timor-Leste sites MK2 and Laili was not as intensive as at coastal sites throughout Wallacea such as Asitau Kuru (O'Connor et al., 2011), Makpan (Kealy et al., 2020), Tron Bon Lei (Samper Carro et al., 2016) and Here Sorot Entapa (O'Connor et al., 2019) (Figure 1). Inland fishing therefore, would have been a complementary component of terrestrial subsistence strategies that may have focused on gathered plant foods and even arboriculture until the spread of neolithic farming societies (O'Connor, 2015). This is a pattern reflected at inland sites elsewhere in Southeast Asia such as Niah cave on Borneo west of Wallace's line, where small amounts of freshwater fish complement freshwater shellfish gathering and hunting of medium to large forest mammals (Piper and Rabett, 2009). East of Wallace's line on the large Wallacean island of Sulawesi, medium to large endemic mammals were also hunted and freshwater shellfish gathered (Glover, 1981; Simons and Bulbeck, 2004). Freshwater fish and eels are rare in archaeological assemblages on the island. They have been identified at the inland shelter Gua Mo'o hono, but are confined to the late Holocene levels (O'Connor et al., 2018).

There are differences in fishing behavior over time between MK2 and Laili. Fishing activity was concentrated during different periods at each site. At Laili the majority of the fish remains were located in the lower part of the stratigraphy in the initial Pleistocene period of settlement ~44.6 kya (63.1%), although fishing continued into the terminal Pleistocene period with the same diversity of foraged environments (36.9%) after the LGM (Table 1). At MK2 most of the fishing was conducted during the Holocene period, possibly reflecting warmer wetter conditions and elevated freshwater levels in the lake, as well as more intensive human occupation in Timor-Leste after the LGM (Hawkins et al., 2017; Shipton et al., 2019, 2021). There were very few fish bones in the MK2 pleistocene deposits, with 92.79% of the fishing being identified in the early Holocene and late Holocene deposit from 11,000 years ago to the modern era, which could be a function of increased distance from the coastline between 40,000 and 31,000 BP (Table 1). These spatio-temporal differences in fishing practices could also be due to differences in regional culture as well as the effects of climate change on different habitats. Laili was perhaps occupied while procuring lithic resources from the river bed (Hawkins et al., 2017), while MK2 was likely occupied as a basecamp from which a variety of activities were undertaken during periods of high lake levels in Ira Lalaro (Veth et al., 2005; Langley and O'Connor, 2015, 2019; Langley et al., 2016), which probably requiring scheduling. Coral reef associated taxa at Laili indicate that reef platforms were present near the north coast of the site despite the steep coastline (Hawkins et al., 2017). MK2 shares similarities with Asitau Kuru, which are both situated near the eastern tip of Timor. Both sites had limited occupation during the LGM, likely due to sharp reduction in rainfall during the LGM which made settlement less favorable (O'Connor et al., 2011; Hawkins et al., 2017; Shipton et al., 2019; Roberts et al., 2020). MK2 and Asitau Kuru also had a shared material culture history that included shell beads and ochre pieces, also found at the neighboring sites of Matja Kuru 1 and Lene Hara respectively (Langley and O'Connor, 2015; Langley et al., 2016; O'Connor et al., 2019). Further, the presence of Scombridae (mackerels, tunas, bonitos) recorded at MK2 has resonances with Asitau Kuru, highlighting a possible connection between the two sites of a highly specialized fishing tradition targeting this fast-moving predator off the edge of the reef.

These findings demonstrate that with access to more adequate modern reference specimens it is possible to identify the different habitats of archaeological fish specimens by comparative morphological analyses and thus reveal a greater range of H. sapiens adaptations to the tropical islands of Wallacea. Our results allow us to hypothetize that H. sapiens in Timor-Leste were able to utilize a flexible and sophisticated set of fishing strategies across a diverse range of inland and coastal aquatic habitats from at least 45 kya. Such foraging adaptations, requiring the scheduling of subsistence activities between the coast and hinterland wetland habitats, would have required a high degree of planning as well as diverse capture technology. This mobility is likely a precursor to the development of trade and exchange networks by the terminal Pleistocene in the Wallacea (Reepmeyer et al., 2011, 2019).

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

CB: conceptualization, formal analysis, investigation, methodology, writing—original draft, and writing—review and editing. SH: conceptualization, writing—original draft, and writing—review and editing. CS: investigation, project administration, supervision, and writing—review and editing. TI and A-MS: conceptualization and writing—review and editing. SSC: formal analysis and writing—review and editing. SO'C: funding acquisition, investigation, project administration, supervision, and writing—original draft. All authors contributed to the article and approved the submitted version.