Supporting Information Masaki Fujita et al. SI Text 1. Stratigraphy, dates, and periods of human occupation at Sakitari Cave The excavated sequence at Trench I and Pit 1 show inclination of about 20° toward the cave’s interior with no signs of substantial water flow or talus cone (Fig. S2), indicating dominantly gradual, colluvial sedimentary process from the outside of the cave. Immediately below the top soil, ~30-cm thick, multi-stratified flowstone (FS) containing shards of Jomon potteries spread all over the Trench I. This FS layer is dated to between 2,800 and 11,000 cal BP, suggesting multiple episodes of flowstone formation and erosion during the early−middle Holocene period. The clayey Pleistocene strata were protected beneath this FS layer. We recognized three major units within the terminal Pleistocene sequence, Layer I, II, and III from top down. All these units include charcoal fragments, but Layer II is distinguished from Layer I and III primarily by significantly higher density of charcoals and the resultant dark color. Layer II is also less compact compared to Layer I and the uppermost part of Layer III. Layer II was further subdivided into two charcoal belts (II-1B and II-2), and the sediments above (II-1A) and below (II-1C) the upper charcoal belt (II-1B)(Fig. 2, Fig. S2). Layer III was also subdivided into the upper and the lower parts by the boundary drawn about 2 m below the ground surface, primarily because the remains of two species of extinct deer were found only from the latter levels. AMS radiocarbon dating shows that Layer I is 13,000−16,500 cal BP, Layer II is 20,000−23,000 cal BP, and Layer III is 23,500−25,000 cal BP (the upper part) and 26,000−36,500 cal BP (the lower part) (Fig. 3, Table S1). The dates for the 39 samples are consistent with the stratigraphy, indicating little post-depositional disturbance of the terminal Pleistocene strata. A single marine shell sample (PLD-19424) from Layer I returned an unexpectedly old age (27,300 cal BP) for that layer. This may be an instance of 1 an upworked older shell, or otherwise people collected an old shell for some use. A charcoal (PLD-24993) and a freshwater snail (PLD-27744) which show relatively young ages obtained from for the layer III probably were intruded from layer II. Identified wood charcoal samples include genus Camellia and possibly Symplocaceae. The dates obtained for these samples may be older than the depositional dates by as much as their life durations of several hundreds to a thousand years. Other samples may have been variously affected by old carbon from the karst. Still, the remarkable stratigraphic consistency of the obtained dates suggest that such effects are minimal, if any. The terminal Pleistocene layers (as well as the Holocene FS layer) offer various evidence of human activities such as charred animal remains, dense charcoals, humanly transported marine shells, shell and stone artifacts, as well as human remains (Fig. 4). Although there appear to be small chronological gaps between the dated samples from Layer I and II (3,500 years) and Layer II and III (2,000 years), the evidence of similar largely seasonal use of this cave throughout the sequence suggests continuous occupation of Palaeolithic people on the Okinawa Island throughout the terminal Pleistocene. The timing when people began to use this cave is a question we cannot answer at the present stage of the research. An atlas and rib of a human infant individual were found at Pit 1, from the lower part of Layer III. AMS ages for three charcoal and freshwater snail samples collected from the vicinity of the human remains are 28,700−31,400 cal BP (1σ ranges for the three samples combined), and humans were clearly at this cave by this time. About 10 cm below the human level, eleven isolated cervid bones including a charred lumber vertebra, as well as freshwater crabs and snails, some of which are also charred, were found together with a charcoal fragment dated to 36,500 cal BP (PLD-16469 in table S1). 2. Stone artifacts The terminal Pleistocene strata of Sakitari Cave have so far yielded only a few stone artifacts: three quartz flakes from Layer I and a possible grindstone made of sandstone from Layer II. Both stones are not available locally and must have been transported into the cave 2 by humans. The quartz flakes are tiny (<3 cm in length) and amorphous (Fig. 4). They bear patterns of breaks most similar to experimentally knapped quartz (1, 2), but their small sizes and amorphous shapes make it difficult to infer their function. Grindstones and/or retoucher must have been necessary to manufacture the fishhooks and scrapers (Fig. 5, Fig. S8). A small sandstone which have smoothed surface probably was used for these purposes. High-quality lithic raw materials such as siliceous stones are unavailable in the karstic areas of southern Okinawa Island where the Sakitari Cave is located. The nearest outcrop of quartz is 30 km away from the cave and that of chert is 50 km away. We hypothesize that this explains the general paucity and crudeness of stone tools from Okinawa; Palaeolithic people here probably relied on other types of raw materials such as marine shells described below and in the main text. 3. Bivalve shell tools Morphotypes The terminal Pleistocene marine bivalve assemblage from Sakitari Cave is mainly represented by Veneridae and Septifer bilocularis (Table S2). The Veneridae includes larger species (Callista chinensis and Meretrix sp. cf. lusoria) that were used as tools, and smaller species (Sunetta kirai) perforated to create beads. A comparatively small bivalve species, S. bilocularis, was also used as tools. These bivalve tools are described below. Table S3 lists the larger Veneridae bivalves so far recovered from the terminal Pleistocene strata of Sakitari Cave. All of these 22 specimens are fragmented. Many of them bear small flaking scars indicative of retouch and/or use-wear, and thus are identified as tools, while others may include debris from the tool manufacturing (Figs. S3 and S5−S7). We classified these Veneridae fragments into the following three morphotypes: 1) fragments with hinge, 2) trapezoid form, and 3) other amorphous fragments. The longer margin of a trapezoid form piece consists of (largely) intact, convex ventral margin of the shell (the margin opposite to the umbo). The shorter margin is a broken, often concave, retouched edge on its dorsal side (near the umbo). Several tiny shell chips were also found from Layer II, suggesting that the shell tools were manufactured at this cave. 3 The trapezoid form dominates the Sakitari Cave fragmented Veneridae shell assemblage: seven specimens were found from Layer II and another from the uppermost part of Layer III (Figs. 4 and 5). According to an experimental study (3), their manufacturing process is as follows: First, a shell is fragmented by dealing a blow to the center of its external surface. Next, a trapezoidal piece is selected from the broken fragments. Then, its fractured, shorter margin (on the dorsal side) is retouched from the external surface of the shell to create a concave margin. In contrast to the fragmented conditions of these larger and thicker Veneridae, the smaller S. bilocularis bivalves are more or less complete (Fig. 5D). Two of the three S. bilocularis specimens bear possible traces of use on their ventral margins as explained below (Fig. S7). Use-wear analysis The use-wear features on the modern experimental Veneridae shells (Fig. S4) were compared to the fragmented bivalves from Sakitari Cave (Table S3 and Figs. S5−S7). Out of the eight trapezoidal Veneridae fragments, the following three specimens from Layer II bear use-wear traces on their short, concave margins (Fig. S5). No. 4 shows a polished micro-plane and striations running perpendicularly to the edge. The concave morphology of the working edge and the above use-wear pattern suggest that this piece was used to scrape bar-like objects. The polish exhibits bright, smooth, flat appearance similar to the polish type observed on the experimental specimens used for bamboo-scraping (fig. S4A). The polish resulted from wood-scraping also exhibited comparable features with that from bamboo-scraping. Hide-scraping produced a matte, rounded polish surface on the edge (Fig. S4B), a pattern identical to the use-wear on lithic specimens (4). In contrast, scraping an antler or a bone did not develop polish but yielded an abraded surface (Fig. S4C). Accordingly, No. 4 was probably used for scraping bamboo or wood. No. 5 and No. 6 (Fig. S5) also show polish and striations or rounding on their concave edges, suggesting utilization as scrapers, although the polish on these latter pieces is not developed enough to infer the worked materials. 4 Among the other non-trapezoid Veneridae fragments, three pieces showed clear use-wear traces. One of them (No. 1 in Fig. S3) shows polish, abrasion, and striations running perpendicular to the retouched edge, suggesting its function as a scraper. The other two artifacts (Nos. 2 and 15 in Fig. S6) exhibit alternate edge-damages, a pattern frequently observed on experimental cutting-tools made of shells (this study) as well as stones (5, 6, 7). In addition, No. 2 shows abrasion, polish, and striations running parallel to the edge on both the external and internal sides of the dorsal margin, supporting cutting activity with this edge. The same formation pattern of polish and striations are also observed on No. 15. The worked materials for the above three specimens could not be assessed because of minimal polish on these shells and the small sample sizes of our experiments.