Taphonomic analysis at Liang Bua reveals the behavioral and technological capabilities of Homo floresiensis

AbstractSince its discovery, Homo floresiensis—an extinct, short-statured, and small-brained hominin species from Flores, Indonesia—has often been ascribed unexpectedly advanced behaviors, such as hunting large game and using fire. Here, we report the results of a systematic taphonomic study sampling the proboscidean bone assemblage at Liang Bua where the frequency and locations of predatory marks, along with skeletal part profiles, show that Komodo dragons likely had primary access to these remains leaving behind only low-utility elements for H. floresiensis to scavenge. Moreover, no signs of intentional use of fire are present in the stratigraphic units associated with H. floresiensis. Together, these results suggest that H. floresiensis was not as behaviorally advanced as originally suggested and provides critical insights into the behavioral repertoire of H. floresiensis, raising important questions about its ancestry. SIGN UP FOR THE AWARD-WINNING SCIENCEADVISER NEWSLETTER The latest news, commentary, and research, free to your inbox daily INTRODUCTIONHomo floresiensis was originally described as having relatively advanced behaviors for a short-statured and small-brained hominin species based on purported evidence of fire use and the hunting of large game (1–3). For example, skeletal remains of H. floresiensis and a dwarfed species of proboscidean (Stegodon florensis insularis) were uncovered together at Liang Bua in association with dense concentrations of stone artifacts, interpreted at the time of discovery as “big game” hunting technology (2, 3) (Fig. 1). This interpretation was further amplified by evidence of cutmarks reported on three Stegodon bones (4, 5) and endocast reconstructions of H. floresiensis suggesting an unusual expansion in the frontal polar region associated with higher cognitive processing (6). Some of the smaller animal remains at the site were also described as charred, implying that they were burned by H. floresiensis, supporting the idea of a relatively small-brained yet behaviorally advanced fire-using hominin (2, 3). Because hunting large game and controlling fire is generally associated with large-brained hominins, such as Neanderthals and modern humans (7–9), the association of these behaviors with H. floresiensis was, and continues to be, particularly unexpected.

However, systematic taphonomic and zooarchaeological analyses of the Liang Bua Stegodon assemblage are needed to fully understand the extent of H. floresiensis’ dietary strategies and potential use of pyrotechnology.Fig. 1. Site location and schematic of the stratigraphy from Liang Bua.(A) Location of Flores within the Indonesian archipelago (gray: current coastal land masses; blue: coastlines during the last glacial maximum). (B) Location of modern towns (white circles), Liang Bua (red triangle), and other sites (yellow circles) on Flores that have yielded Stegodon remains. Maps generated using QGIS [QGIS Development Team (74)]. (C) Site plan of excavated sectors that yielded the skeletal samples used in this study. (D) Stratigraphic composite with associated age ranges modified from (13). Subunits from which murine (green) and Stegodon (blue) remains were sampled in this study are delineated on the right.We used taphonomic and zooarchaeological analyses (see Materials and Methods) to test whether H. floresiensis used fire and hunted Stegodon at Liang Bua. To determine the order of predator access to carcasses on a paleolandscape, comparative data such as mark locations and frequencies on prey long bones as well as prey element profiles are required (10, 11). However, such data derive almost exclusively from the prey of mammalian predators and do not apply to the ecological context at Liang Bua, in which the only predator H. floresiensis would have competed with for access to Stegodon was Varanus komodoensis (Komodo dragon), the largest extant reptile on earth. Unlike mammals, Komodo dragons have serrated ziphodont dentitions that create a distinctive, complex patterning to the bone surfaces of their prey. We conducted a controlled feeding experiment at Zoo Atlanta to generate a sample of Komodo dragon tooth marks and compared these with experimentally generated cutmarks.

We then randomly sampled 3155 Stegodon bone fragments (~27% of total assemblage) from stratigraphic Unit 1 [~190 to 60 thousand years ago (ka)] and Unit 2 (~60 to 50 ka), both of which are only associated with H. floresiensis (12, 13), as well as 6906 murine rodent skeletal elements from stratigraphic Units 1 and 8, the latter of which is <11 ka and only associated with H. sapiens (Fig. 1) (see Materials and Methods). All 10,061 elements were examined for evidence of exposure to fire, and the Stegodon sample was further analyzed to identify the probable agent(s) responsible for the accumulation based on bone surface modifications and skeletal abundances.RESULTSDocumenting skeletal damage in Komodo dragon preyTo distinguish Komodo dragon tooth marks from cutmarks in the Stegodon assemblage, our controlled feeding experiment used a prepared adult goat carcass, with head and distal appendages removed, that was fed to a captive Komodo dragon. All remaining 72 bone elements were analyzed for tooth marks after feeding. Twenty-six of these elements had a total of 192 tooth marks with a mean of 7.4 marks per element (range: 1 to 55) with a majority found on the upper forelimb (42%, humeri; 14%, scapulae) (data S1). Scores made up ~95% of these marks, followed by pits (2.5%), notches, hooks, and furrows (<1% each). Twenty-seven scores contained internal microstriations, 2 had external microstriations, and 14 had asymmetrical microstriations emanating from the main groove to form a “fan” (Fig. 2, figs. S1 and S2, and table S1). These “fans” likely resulted when the posterior serrated edge of the tooth made contact with the bone during consumption. Score shape also varied depending on the type of bone and was likely the result of varied grip and rip actions (fig. S2 and movie S1). For example, scores found primarily on long bone shafts were typically shallower and wider compared to relatively deeper and narrower marks found on flatter and angular bones like ribs, vertebrae, and scapulae.

In addition, depth was either consistent throughout the length of the score or varied in the form of a “pit and drag” effect. These results—in combination with those from a previous controlled feeding experiment involving Komodo dragons and multiple feeding sessions (14)—suggest that Komodo dragon tooth mark frequencies are highly variable but are consistently located on skeletal elements with “substantial” amounts of flesh.Fig. 2. Models of tooth marks created by Komodo dragons.(A to F) Scores of Komodo dragon tooth marks generated from our experimental study.A three-dimensional (3D) quadratic discriminant analysis (QDA) using a resubstitution model classified the experimentally generated Komodo dragon tooth marks (n = 72), cutmarks (n = 403), and trample marks (n = 130) with 84% correct classification (Fig. 3 and table S2). The canonical scores between groups were statistically significant (P < 0.0001) (fig. S3). When plotted in canonical space, the Komodo dragon tooth marks separated from both cutmarks and trample marks along the negative end of the x axis (can1) (70%) with some overlap with both groups (Fig. 3). Trample marks and cutmarks clustered toward the positive end of can1 but with considerable overlap while separating primarily along the y axis (can2) (30%). Almost all the variables included in the analysis were statistically significant in separating cutmarks from Komodo dragon tooth marks (P < 0.05) apart from the maximum width (figs. S3 and S4). However, the width collected at the deepest point, as well as the depth, angle, and roughness, contributed the greatest for distinguishing Komodo dragon tooth marks from cutmarks and trample marks. Overall, Komodo dragon tooth marks tend to be shallower, shorter, and have a wider profile angle compared to cutmarks (fig. S5).Fig. 3. QDA of marks from the Stegodon assemblage and experimental samples.Analysis includes 3D variables plotted in canonical dimensions comparing trample marks (trample), cutmarks (CM), and Komodo dragon tooth marks (KTM) with marks from the Liang Bua Stegodon assemblage. Experimental samples are shown in yellow (trample), blue (cutmarks), and red (Komodo dragon tooth marks).

The black icons represent identifications for the Liang Bua Stegodon sample. Liang Bua marks are labeled on the basis of high-confidence identifications, qualitative features, and posterior probability scores.Results from the Stegodon assemblage at Liang BuaAmong the Stegodon sample, 70% [709 subadult and 1511 adult bones and an additional 712 number of identified specimens (NISP)] were identified to elements but only 22% were sided to either left or right, enabling calculation of the minimum number of elements (MNE) (Fig. 4 and Table 1). On the basis of the most frequently represented skeletal elements, we calculated a minimum number of individuals (MNI) of nine subadults (unfused basiocciputs) and two adults (patellae), consistent with previously estimated demographic age profiles (5). Fragmentation rates (MNE/NISP) indicate that adult elements (0.06) were more fragmented compared to those of subadults (0.21). A large proportion of unidentifiable bone fragments was represented (29.8%), further demonstrating the high rate of fragmentation, most of which was due to postdepositional processes (60%) and recent breakage (28%) (table S3).Fig. 4. Size range and frequency of Stegodon skeletal elements.Boxplots of elements ordered by median values. Points are colored according to age-at-death (i.e., adult or subadult) with a density plot of each group displayed on the right. For the boxplots, center line is the median, upper and lower hinges are the first and third quartiles (25th and 75th percentiles), whiskers are 1.5 times the interquartile range, and points are all data points.Age (MNI)Subadult (8.3)Adult (1.9) ElementNISPMNEMAU%RANISPMNEMAU%RATotal