Adaptations to sea level change and transitions to agriculture at Khao Toh Chong rockshelter, Peninsular Thailand

This study reports on an analysis of human adaptations to sea level changes in the tropical monsoonal environment of Peninsula Thailand. We excavated Khao Toh Chong rockshelter in Krabi and recorded archaeological deposits spanning the last 13,000 years. A suite of geoarchaeological methods suggest largely uninterrupted deposition, against a backdrop of geological data that show major changes in sea levels. Although there is a small assemblage of mostly undiagnostic ceramics and stone artefacts, there are some distinct changes in stone artefact technology and ceramic fabric. There is a substantial faunal assemblage, with changes in both the mammalian and shellfish taxa during the Pleistocene-Holocene transition that correlate with local sea level fluctuation. This assemblage provides an opportunity to explore subsistence behaviours leading up to the transition to the Neolithic. We explore the implications for current debates on the prehistoric origins of agricultural subsistence in mainland Southeast Asia. The data highlight the importance of local contingencies in understanding the mechanisms of change from foragers to agriculturalists. Disciplines Medicine and Health Sciences | Social and Behavioral Sciences Publication Details Marwick, B., Van Vlack, H. G., Conrad, C., Shoocongdej, R., Thongcharoenchaikit, C. & Kwak, S. (2017). Adaptations to sea level change and transitions to agriculture at Khao Toh Chong rockshelter, Peninsular Thailand. Journal of Archaeological Science, 77 94-108. Authors Ben Marwick, Hannah G. Van Vlack, Cyler Conrad, Rasmi Shoocongdej, Cholawit Thongcharoenchaikit, and Seungki Kwak This journal article is available at Research Online: https://ro.uow.edu.au/smhpapers/4365 Adaptations to sea level change and transitions to 1 agriculture at Khao Toh Chong rockshelter, Peninsula 2 Thailand 3 Ben Marwick (University of Washington, University of Wollongong, bmarwick@uow.edu.au) 4 Hannah Van Vlack (San Jose State University) 5 Cyler Conrad (University of New Mexico) 6 Rasmi Shoocongdej (Silpakorn University) 7 Cholawit Thongcharoenchaikit (National Science Museum of Thailand) 8 Seungki Kwak (University of Washington) 9

Holocene (c. 6-3.5 k BP) --is sparsely represented in the archaeological record. Southeast Asia 39 has a rich and well-documented archaeological record for the later Holocene, when people were 40 living more sedentary lifestyles, for example at Khok Phanom Di in Thailand and Man Bac in 41 Vietnam (Higham and Bannanurang 1991;Oxenham et al. 2011). There are also many cave and 42 rockshelter sites representing Pleistocene forager lifestyles, such as Tham Lod in Thailand and 43 Xom Trai in Vietnam (Shoocongdej 2006;Moser 2001). 44 However, during the middle Holocene, the archaeological record in mainland SEA is particularly 45 sparse. This gap in archaeological evidence for the region has been called "the missing 46 millennia" (White and Bouasisengpaseuth 2008:39). It is an important period because major 47 changes occurred during this time. Ceramics appeared in many parts of Southeast Asia; 48 domesticated plants such as millet and rice appeared; stone artefact technologies transitioned 49 from mostly flaked to mostly ground stone artefacts; and settlements expanded from primarily 50 karstic upland and estuarine landscapes during the early Holocene to include inland alluvial 51 lowland villages by the late Holocene (White 2011). But the sparse representation of this period 52 in the archaeological record means that questions of the timing and character of these changes 53 remain difficult to answer. 54 In this paper we present evidence of human activity from coastal Thailand that spans "the 55 missing millennia." Khao Toh Chong rockshelter is significant because it has a rich faunal record 56 spanning the middle Holocene, and is located in an area with a relatively detailed history of 57 regional sea level change. This provides a unique opportunity to investigate locally contingent 58 factors, such as the effect of sea level changes on human subsistence behaviours during the 59 transition from forager to agricultural economies. We report on a geoarchaeological analysis of 60 the site to provide a local environmental context of the human occupation. This analysis also aids 61 our understanding of site formation processes and artefact taphonomy. behaviours has been proposed to be a direct result of the transition to agriculture (White 1995), 66 and is evident in surrounding regions. The Guangxi Province of southern China has extensive 67 evidence of a forager economy with a semi-sedentary lifestyle during c. 7-4 k BP (Higham 68 2013). Cave occupation continues until 6 k BP in Xianrendong and 5-4 k BP in Zengpiyan, and 69 more than 30 open sites containing shell middens have been found on the terraces of the 70 Zuojiang, Youjiang and Yongjiang rivers near Nanning, in southern Guangxi (Chi and Hung 71 2012; Fu 2002). Occupation of these sites, characterized by the largest, Dingsishan, spans 10-5.5 72 k BP. The sites include pottery manufacturing workshops, cemeteries and large quantities of 73 aquatic and terrestrial animal bones, indicating that fishing and hunting were important activities 74 (no cultivars have been recovered). The archaeology of this region gives the impression of a 75 continuous sequence of human occupation. We see gradual, overlapping adaptations resulting in 76 changes in landscape use, the appearance of pottery and use of cemeteries, and at a much later 77 date an agricultural economy. The pottery and burial practices of the Dingsishan shell middens 78 are identical to those found at the Da But sites of northern Vietnam, such as Da But, Con Co 79 Ngua, Ban Ban Thuy, Lang Cong and Go Trung (Viet 2007). These sites were occupied by 80 hunter-gatherer populations during 7.5-4 k BP (Viet 2007). Polished axes, pestles and mortars 81 suggest cultivation, but clear evidence of food production only appears around 3.8-3.5 k BP at 82 sites such as Man Bac with domesticated pig remains (Sawada et al., 2011). 83 While this gives a picture of continuity between hunter-gatherers and agriculturalists in southern 84 China This landscape has been exposed to major changes as sea levels rose and fell during the Late 106 Pleistocene and Early Holocene (Voris 2000;Sinsakul 1992 to a maximum of 2 m amsl at 2.5 k BP. Regression continued from that time until the present sea 119 levels were reached at 1.5 k BP. 120 The evidence for these sea level changes comes from direct dating of marine shells and peat 121 deposits at geological sites in peninsular Thailand (Sinsakul 1992). Tjia (1996)  faunal remains including marine shellfish dating between 6 k and 5 k BP (Srisuchat and 145 Srisuchat 1992). Pak Om has a dense and diverse archaeological deposit, but its two dates of 146 9.35 k and 3.01 k BP come from the same layer, so the chronology is uncertain (Srisuchat 1997).

147
Khao Tau in Pang Nga is a site complex with deep stratification and abundant cultural materials 148 dating to 5.25 k and 4.75 k BP (Srisuchat and Srisuchat 1992). Finally, there is the Tham Sua 149 shell midden in Krabi that is a deposit of marine shell greater than one meter deep and with a 150 basal date of 6.44 k BP (Anderson 2005). 151 These previous excavations demonstrate human occupation at several sites in peninsular 152 Thailand during the critical time of sea level changes in the Holocene. However, the level of 153 available detail at these sites provides neither a clear picture of stratigraphic integrity, nor their 154 subsistence behaviour. The goal of our work at Khao Toh Chong was to build on this previous 155 research by analysing an assemblage spanning the Holocene, and by conducting 156 geoarchaeological analyses at the site to assess stratigraphic integrity and provide local 157 environmental context of the human occupation. 158 Bulk sediment samples were collected from a column taken from the south wall of excavation 205 trench A. Sub-samples of sediment (1 g) from each context were individually dried at 60°C for 206 24 hours for particle size analysis. These sub-samples were sieved to remove the >2 mm 207 particles, and the carbonates were removed by washing the sample in 20 mL of 1 M HCl. 208 Samples were then centrifuged and treated with 30 mL of 30% H 2 O 2 for an hour to remove 209 organics (Scott-Jackson and Walkington 2005). Additional drying occurred for 30 hours in a 210 60°C oven. Each sample was added to a mixture of deionized water and surfactant Triton X 10 211 and agitated before being run in a Horiba LA-950 at the University of Washington (UW) 212 Materials Science Department. A quartz refraction index of 1.458 was used during analysis and 213 the R package G2Sd v2.1.5 was used to compute summary statistics (Fournier et al. 2014). 214 We measured pH and EC using a portable Oakton Waterproof Dual Parameter PCSTestr 35 on 215 sub-samples with a 1:1 ratio of sediment to deionized water. Soil organic material (SOM) and 216 calcium carbonate content were measured by the Loss on Ignition method (Gale and Hoare 217 1991), as the percent of mass lost after heating samples to 600ºC for 4 hours and 1000ºC for 2 218 hours. Magnetic susceptibility was measured using a Bartington MS2 Magnetic Susceptibility 219 Meter with 10 cm 3 of sediment analyzed in sample pots at low and high frequency following 220 Dearing (1999). Three replicates for each sample measurement of low and high frequency 221 susceptibility were taken following Gale and Hoare (1991). 222 Organic carbon isotopes were analysed by sub-sampling 2 g of sediment which was dried at 223 60°C for 24 hours, then sieved to remove the >2 mm particle size fraction (Hartman 2011), and 224 macro-organics were manually picked out and discarded. After sieving the samples were ground 225 for 5 minutes using a mortar and pestle. Mineral carbonates were removed by placing the 226 samples in 60 mL of 1 mol HCl for 24 hours, stirring every 10 hours of the 24 hour period 227 (Millwood and Boutton, 1998). The HCl was rinsed from the samples by adding 60 mL of 228 deionized water into the samples for one minute and then drying at 60°C for 48 hours; this step 229 was repeated three times. Isotope measurements were conducted using a Costech Elemental 230 Analyzer, Conflo III, MAT253 at the UW Earth and Space Sciences IsoLab. 231 For XRD analysis, following McGrath et al. (2008), we sub-sampled 2 g of >2 mm sediment and 232 ground it to a fine powder. Next 20 mL of 30% H 2 O 2 was used to remove organic matter. After 233 effervescence, sediment samples were dried for another 60°C for 24 hours. Samples were ground 234 again, then scanned on a Bruker D8 Focus X-ray Diffractometer from 5° to 75° 2θ with a Cu 235 radiation source at resolution 0.02° steps per second with 40 kV and 40 mA power output. MDI 236 Jade 9 software was used to identify minerals. 237 For compositional analysis by ICP-AES a 1 g sub-sample of sediment was prepared with an acid 238 digest extraction, following Misarti et al. (2011). The sample was added to 10 mL of HNO 3 and 239 heated at 90°C for 15 minutes. Another 5 mL of HNO 3 was next added and heated at 90°C for 60 240 minutes. Next, deionized water, 30% H 2 O 2 and 10 mL HCl were added and heated for 60 241 minutes. The samples were then diluted with deionized water and filtered before ICP-AES 242 analysis. This acid digest provides a broad spectrum of elements in a known volumetric 243 concentration, suitable for ICP-AES analysis (Balcerzak, 2002;Carter, 1993

278
The key findings from our field observations during the exacavation were that the faunal 279 assemblage was deposited with relatively few macroscopic traces of post-depositional 280 disturbance ( Figure 3). We did not encounter any human burials or animal burrows and there was 281 very limited termite activity visible in the deposit. We did not reach bedrock, or sterile deposits, 282 due to time constraints. All excavated materials are currently stored at the Silpakorn University 283 Faculty of Archaeology's Phetchaburi campus. 284 Five charcoal samples and five shell samples returned radiocarbon age determinations (Table 1). 296 The ages of these shells are offset from the ages of the charcoal by an average of 2945 years, 297 indicating a substantial reservoir effect. Considering only the charcoal dates, the excavated 298 deposit spans from before 13.5 k cal. BP through to about 0.15 k cal. BP ( Figure 5). 299 The depth-age relationship for the dated samples is strongly linear, suggesting a constant rate of 300 sediment accumulation (Figure 4). Although there is nearly a meter between the lowest and 301 second lowest charcoal samples, the linear tendency of the shell samples that span this gap 302 suggest that the accumulation of sediment at the site has been constant through the Holocene. 303 Using the ages of the charcoal samples, we computed a non-parametric chronology model to 304 estimate the approximate ages of undated excavation units. Using this model, we estimate the 305 date of the lowest excavation level to be approximately 16.8 k cal. BP.

Chemical analyses and magnetic susceptibility 324
The results of the chemical, magnetic susceptibility and particle size analyses are depicted in 325 Figure 6. The pH values at KTC are strongly alkaline throughout, with a shift occurring from pH 326 9.1 to 7.6 between contexts 5 and 6 of trench A (0.4-0.53 m below surface). Electrical 327 conductivity (as a proxy for soluble minerals) and soil organic matter decline sharply below the 328 surface, probably due to natural decay of organics. Soil carbonates are steady between 8% and 329 12% throughout, reflecting a continuous contribution from the limestone rock of the shelter. Low 330 frequency magnetic susceptibility peaks in context 5 of trench A (0.40 m below surface), 331 indicates an enrichment of magnetic minerals in the deposit. Context 5 has the highest proportion 332 of carbonates (12%), which would reduce magnetic susceptibility; the change in this context is 333 not a simple dilution of magnetic minerals by diamagnetic minerals. 334

Carbon isotope analysis 335
The δ 13 C values at KTC range between -28.75‰ and -26.2‰, with values becoming increasingly 336 depleted in more recent times ( Figure 6). The tissues of C 3 plants have δ 13 C values ranging from 337 −32‰ to −20‰, while those of C 4 plants range from −17‰ to −9‰ (Deines 1980    Glaser 2004). The majority of these elements are strongly positively correlated (Table 5), and 363 there are no significant negative correlations. 364    However, disturbance is not a significant factor at KTC as supported by the mineralogical and 389 sediment particle size data. Similar depositional processes occurred at Spirit Cave in northern 390 Thailand (Gorman 1970). For example, radiocarbon dating of residues on ceramics from Spirit 391 Cave obtained much younger dates (c. 3 k BP) than the stratigraphically associated charcoal 392 samples (c. 7.6 k BP; Lampert et al. 2003). This shows that there is probably some mixing in the 393 stratigraphic layers at Spirit Cave. Comparatively, the KTC ceramics may have also shifted 394 vertically over time due to the episodes of regional increases in precipitation from either the 395 water table or seasonal monsoonal storms. 396 The archaeological sequence at KTC shows signs of change over time, similar to the 397 geoarchaeological sequence described above, indicating that disturbance has not been so 398 extensive as to completely erase time-ordering of artefacts in the deposits. The stone artefact 399 technology changes from to large flaked cores and flakes made from coarse-grained 400 metamorphic rock in the lower levels to polished adze flakes made from finer-grained rock in the 401 upper levels. The ceramic assemblage also changes from thick, red sherds with frequent incised 402 decorations in the lower levels to predominantly black sherds in the upper levels. However, the 403 small number of artefacts in the deposit overall limits the degree to which we can distinguish 404 these changes as part of a major regional trend or idiosyncratic use of this site. 405   Rattus remotus 0 (0) 0 (0) Cyclophorus cf.  Of the identified invertebrates, nine taxa were identified to the species level while an additional 430 fourteen were identified to a broader degree of taxonomy ( Ecological indices of taxonomic diversity and evenness vary over time, suggesting complexities 439 in forager behaviour (Table 9). Generally, these indices have low values, indicating both low 440 diversity and the dominance of a small number of taxa in the assemblage. This is largely 441 controlled by the abundance of Neoradina prasongi, which dominate the assemblage in the lower 442 levels despite a greater number of other taxa also present. In the upper levels where Neoradina  443 prasongi is absent, the diversity and evenness indices increase but overall counts are low 444 suggesting the site was less frequently used for subsistence activities. Visual inspection of the stratigraphic plot of the KTC data ( Figure 6) suggests that the magnetic 457 susceptibility frequency dependency values of track mean particle size more closely than they 458 track low frequency magnetic susceptibility. This indicates that soil formation and weathering 459 processes control magnetic susceptibility more than burning processes, such as cooking, at the 460 site (Dearing et al. 1996 with soil organic matter in the KTC deposits (Table 2). A negative correlation can be explained 464 by a negligible contribution from in situ pedogenesis toward enriching magnetic susceptibility. 465 This suggests that the enhancement of susceptibility may have occured off-site, rather than 466 through in situ processes in the deposit. If the magnetic susceptibility signal is not coupled to 467 anthropogenic burning at the site, as suggested by the the relationship between mean particle size 468 and frequency dependency, the high susceptibility values at 0.40 m below surface (c. 4-5 k cal. 469 BP) may indicate warmer/wetter conditions. One possible mechanism linking higher sediment 470 magnetic susceptibility values to warmer/wetter conditions has been suggested by Ellwood et al. 471 (1997). They propose that higher magnetic susceptibility values might result from increased 472 production of maghemite due to higher pedogenetic rates on the landscape, with enriched 473 sediments washing into and forming site deposits. At KTC see signals of increased site use 474 through artefact discard rates, peaking in contexts 4 and 5. If the mechanism of Ellwood et al.

475
(1997) is plausible, this increase in site use may reflect people seeking shelter during 476 warmer/wetter conditions. Further analyses with remanence (e.g. HIRM, SIRM) measurements 477 will improve our understanding of these relationships. 478 Carbon isotope values indicate a consistent dominance of C 3 plants in the local environment over 479 time, similar to the present-day environment. The small monotonic depletion in carbon isotope 480 values throughout the Holocene suggests that the deposit has some stratigraphic integrity, despite 481 the anomalously deep finds of ceramics. The depletion in carbon isotope values may be due to 482 several factors, including changes in the ratio of C 3 and C 4 plants on the landscape, changes in 483 the growing conditions of plants (such as canopy structure, and water or nutrient stress), changes 484 in the ratios of isotopically distinct organic fractions in the sediment organic matter, and changes 485 in organic inputs from microorganisms in soils (Tieszen 1991 Jackson 2000). Since SOM is probably not the primary driver of δ 13 C values at KTC, then we 492 may be observing a decrease in the relative ratios of C 4 /C 3 plants on the landscape, indicating 493 increasingly dry conditions in more recent periods. dry/hot environments, and high light intensities because C 4 plants are more efficient than C 3 498 species in their use of water, light, and nitrogen (Sage 1999;Pagani et al. 1999). This means that 499 C 3 plants are favored over C 4 plants at times of lower temperature and winter precipitation or 500 during periods of decreased East Asian summer monsoon strength. In the upper 0.2 m, around 3-501 2 k cal BP, at KTC we see increasingly depleted δ 13 C values, suggesting a reduction in C 4 plants 502 as a result of cooler and dryer conditions relative to the Early Holocene. This is consistent with 503 cooler/dryer conditions indicated by a decrease in magnetic susceptibility occurring at KTC at 504 the same time. However, the trend in δ 13 C values at KTC is relatively low magnitude, and 505 isotopic fractionation and microbial activity cannot be fully dismissed as contributing factors 506 (Lerch et al. 2011;Schweizer et al. 1999;Tieszen 1991;Wynn 2007). Carbon isotope values of 507 leaf wax n-alkanes may help to overcome these ambiguities because these are more diagnostic 508 than those from bulk sediments, which contain materials of both terrestrial and aquatic origin. 509 The magnetic susceptibility and carbon isotope data indicate a transition from warmer/wetter 510 conditions at 5-4 k cal. BP to dryer conditions around 3-2 k cal. BP. There are very few nearby 511 comparable records spanning this period, but our interpretations are consisitent with a strong 512 Asian summer monsoon in the Early Holocene, and weakening into the Middle and Late 513 Holocene (Cook and Jones 2012). Lake sediment sequences from northeast Thailand indicate 514 peak Holocene wetness slightly earlier than KTC, at around 7 k and 6.6 k cal. BP, followed by 515 dry conditions between 5.4 k and 4 k cal. BP (Wohlfarth et  The XRD data show variation in the proportion of kaolinite throughout the deposit. The kaolinite 527 is probably derived from the weathering of feldspars and other silicate minerals, and may relate 528 to changes in weathering on the landscape around the site (Nesbitt and Young 1989;Nesbitt et 529 al. 1997). Substantial changes in surface geochemistry are unlikely, due to the absence of 530 correlations between changes in magnetic susceptibility and minerals identified by XRD 531 analysis. If these were correlated, it might suggest episodes of soil formation on the landscape 532 surrounding the site. Thus, we interpret the geoarchaeological data as indicating generally 533 constant conditions over time, rather than resulting from massive large scale bioturbation. 534 The relationships among the elements measured by ICP-AES suggest a single source for the 535 sediments throughout the entire period of deposition. Cluster analysis of the contexts using the 536 elemental data suggests low-level groupings resulting from minor variation (Figure 7). The 537 cluster containing context 1 of trench B, and trench A's contexts 4, 7U and 7L are notable 538 because they are relatively enriched with Ca and Mg, but this is not correlated with carbonates 539 measured by loss on ignition. Overall, the element distributions suggest low variation over time.

540
This homogeneity in the composition of the deposits is consistent with a single source of 541 sediment throughout the history of site formation at KTC. 542 Zooarchaeological assemblage 543 KTC rockshelter has a relatively undisturbed mammalian, reptilian, fish, and molluscan 544 assemblage. Of the taxa recovered at KTC, the riparian fauna is the best indicator of changing 545 forager behavior during the "missing millennia," highlighting the environmental constraints on 546 resource availability. Neoradina prasongi shells constitute the bulk of molluscan food waste in 547 the archaeological assemblage. These gastropods inhabit fresh water stream environments 548 (Brandt 1974), which were likely close in proximity to the rockshelter during this time. Peak 549 discard rates for N. prasongi at KTC occurred at c. 9 k cal BP, suggesting that the most intensive 550 use of the rockshelter for subsistence purposes occurred during the Early Holocene. The 551 abundant turtle or tortoise remains at KTC also suggest that fresh water stream habitats were 552 found near the site. Since KTC was close in proximity to a number of other cave and rockshelter 553 sites with relatively similar chronological and subsistence regimes, it is possible that foragers in 554 this region employed a complex mobility strategy to access fresh water resources and shelter 555 (Brantingham 1991 A decline in freshwater N. prasongi mollusk exploitation occurred in the Holocene, reaching a 558 minimum at 6 k cal BP. Two possibilities may explain this decline; either there is a regional 559 ecological shift from freshwater to mangrove swamp habitats, or changes in the foraging 560 behaviours of prehistoric groups (Shoocongdej 2000(Shoocongdej , 2010. The timing of the lowest amount of 561 shells in the deposit coincides with the peak sea levels, as noted above. Rising sea-levels 562 throughout the Holocene would have shifted mangrove environments closer to the rockshelter 563 over time, which may have influenced the abundance and distribution of locally available 564 resources and freshwater stream environments (Anderson 1990;Horten et al. 2005;Tjia 1996; 565 Sinsakul 1992). These initial faunal data from KTC describe a pattern of forager groups utilizing 566 a diverse range of locally available taxa in the tropical rainforest environment, suggesting that 567 foragers at KTC were able to effectively adapt to shifts in local environmental conditions. 568 Additionally, our radiocarbon dates suggest that the decline in intensive harvesting of N. 569 prasongi during the Middle Holocene may be associated with the emergence of rice agriculture 570 and farming in mainland Southeast (Castillo 2011;Fuller 2011;White et al. 2004). Thus, 571 declines in mollusk utilization may reflect a pattern of rising sea levels. The mechanism here 572 may be a reduction in the availability of suitable mollusk procurement locations, favoring the 573 adoption of agriculture during the Mid and Late Holocene in Peninsular Thailand as a response 574 to these sea level changes. Shell exploitation picks up again at KTC at c. 3 k cal BP, coincident 575 with the regressive phase at 3.7 k to 2.7 k cal. BP described by Sinsakul (1992). This is also 576 when site use changes, with more frequent visits suggested by peaks in the discard of ceramics 577 and lithics. 578 Our data from KTC not only suggest that a subsistence change occurred at the Pleistocene-579 Holocene transition, but that foragers utilizing the rockshelter displayed a pattern of faunal 580 exploitation not widely noted at archaeological sites in Thailand. there is some uncertainty about the effect of bioturbation on artefact distributions. That said, the 610 site is unique because it has not been extensively disturbed by Late Holocene human burials. The 611 faunal assemblage proved the most abundant and interesting aspect of the excavated materials, 612 and broadly confirms some of the patterns previously observed at Lang Rongrien rockshelter and 613 Moh Khiew cave. The foragers occupying KTC practiced a complex strategy of molluscan 614 resource procurement and exploitation. The most striking find is the association between the 615 abundance of shellfish and past sea levels. Low sea levels at the Early Holocene correspond to a 616 peak in shellfish discard, followed by a decline in shellfish and lithic discard at c. 6 k cal. BP, at 617 the same time as the peak Holocene sea levels. 618 There is another small peak in shellfish at c. 3 k cal. BP during a regressive phase, this time 619 accompanied by relatively large amounts of ceramics and lithics. During the Mid Holocene, 620 when the Neoradina prasongi exploitation ceased at KTC, the water table and sea levels were 621 rising while abundances in charcoal (regional fires) became more prevalent (Kealhofer 2003:80; 622 Maloney 1999). During this time, more arboreal taxa were exploited and economic plants begin 623 to appear archaeologically. This faunal discard sequence suggests that local sea levels influenced 624 the intensity of site use. Past human occupants appeared to have found the site favorable for 625 habitation during conditions of low sea levels. Presumably during higher sea levels they sought 626 shelter further inland. In any case, we have shown that adaptation to sea level changes did not 627 require major technological reorganization for the occupants at KTC, but instead was managed 628 by adjusting settlement and land-use patterns to maintain access to resources such as shellfish. The results from KTC confirm the "missing millennia" as a period of important subsistence and 646 technological changes in mainland Southeast Asia. One one hand, we see at KTC a recapitulation 647 of a common sequence in mainland Southeast Asian prehistory. This includes foragers using the 648 site for brief subsistence-related tasks during the Late Pleistocene and Early Holocene, then a 649 transition in the Middle Holocene to people using the site less for foraging activities, but now 650 with ceramics and possibly practicing agriculture, as suggested by the polished adzes. On the 651 other hand, we also see a unique pattern of shellfish exploitation at KTC that is related to the 652 local sea level changes. This relationship highlights the importance of the local context in 653 understanding the mechanisms of change from foragers to agriculturalists. The model proposed 654 by Hunt and Rabett (2014), of a locally contingent protracted process of human modification of 655