Significance

Humans are distinguished from all other primates by their reliance on tool use. When this uniquely human feature began is debated. Evidence of tool use in human ancestors now extends almost 3.3 Ma and becomes prevalent only after 2.6 Ma with the Oldowan. Here, we report a new Oldowan locality (BD 1) that dates prior to 2.6 Ma. These earliest Oldowan tools are distinctive from the 3.3 Ma assemblage and from materials that modern nonhuman primates produce. So, although tool production and use represent a generalized trait of many primates, including human ancestors, the production of Oldowan stone artifacts appears to mark a systematic shift in tool manufacture that occurs at a time of major environmental changes.

Abstract

The manufacture of flaked stone artifacts represents a major milestone in the technology of the human lineage. Although the earliest production of primitive stone tools, predating the genusHomo and emphasizing percussive activities, has been reported at 3.3 million years ago (Ma) from Lomekwi, Kenya, the systematic production of sharp-edged stone tools is unknown before the 2.58–2.55 Ma Oldowan assemblages from Gona, Ethiopia. The organized production of Oldowan stone artifacts is part of a suite of characteristics that is often associated with the adaptive grade shift linked to the genusHomo. Recent discoveries from Ledi-Geraru (LG), Ethiopia, place the first occurrence ofHomo ∼250 thousand years earlier than the Oldowan at Gona. Here, we describe a substantial assemblage of systematically flaked stone tools excavated in situ from a stratigraphically constrained context [Bokol Dora 1, (BD 1) hereafter] at LG bracketed between 2.61 and 2.58 Ma. Although perhaps more primitive in some respects, quantitative analysis suggests the BD 1 assemblage fits more closely with the variability previously described for the Oldowan than with the earlier Lomekwian or with stone tools produced by modern nonhuman primates. These differences suggest that hominin technology is distinctly different from generalized tool use that may be a shared feature of much of the primate lineage. The BD 1 assemblage, near the origin of our genus, provides a link between behavioral adaptations—in the form of flaked stone artifacts—and the biological evolution of our ancestors.

Results

The LG Research Project (LGRP) area is in the lower Awash Valley (LAV), Ethiopia (Fig. 1). The LGRP area contains sedimentary strata that overlie the Hadar (∼3.8–2.9 Ma) and underlie the Busidima (∼2.7–0.16 Ma) Formations (9), representing an interval that is poorly known from the LAV specifically and in eastern Africa generally. Paleoenvironmental and geological analyses have documented a more open habitat than previously recognized in the LAV (9) in association with the LD 350–1 mandible, the earliest hominin fossil with derived mandibular features that define the genusHomo (10).

In 2012, at the BD 1 locality in the eastern LGRP area, a number of stone artifacts were identified embedded in a steep west-facing slope adjacent to the Korkora basalt horst in the Ali Toyta river drainage (Fig. 2). BD 1 is ∼5 km north of the Lee Adoyta region where the LD 350–1 mandible was recovered in 2013 (10). The LGRP excavated 37 m² at BD 1 at an average depth of 1.8 m below surface during field seasons in 2013 and 2015. During this excavation, 300 stone artifacts and 330 associated fossils were recovered in situ (SI Appendix, Table S1) from within a restricted stratigraphic interval at the interface of two distinct sedimentary units (76% of the assemblage comes from a 4–12 cm thick horizon) (Fig. 3).

The bed overlying the archaeological horizon rapidly grades laterally (∼3 m westward) from a clayey-silt with mollusk (gastropod and bivalve) shells to a pebble conglomerate. The underlying clay horizon contains dispersed gastropod shells and displays vertic structure and development of gilgai microrelief. Micromorphological evidence indicates the presence of a relatively stable land surface upon which the artifacts were deposited (SI Appendix, Figs. S1–S3).

The Ali Toyta sedimentary package reflects several phases of lacustrine and perilacustrine deposition followed by subaerial exposure and pedogenic development. The BD 1 archaeological horizon formed on a surface of exposed lacustrine clays following a lake regression with sufficient subaerial exposure to allow for vertisol development. The site was subsequently buried and possibly reworked by low energy sheet flow during a later transgressive phase of the lake that deposited coarser grained material as well as gastropods and bivalves, likely representing a beach or nearshore deposit (Fig. 3 andSI Appendix, Figs. S3 and S4).

Site formation analysis (fabric analysis;SI Appendix, Fig. S5A) and soil micromorphological observations indicate that the artifacts at BD 1 were deposited on an exposed clay surface. The combined presence of small artifacts (<2 cm) and fragile microfossils, the lack of artifact rounding, find orientations (SI Appendix, Fig. S5A), and differences in the spatial distribution of artifacts relative to nonartifactual stones (SI Appendix, Fig. S6) all indicate that the material was not subjected to extensive postdepositional disturbance. Relatively minor artifact reworking subsequent to deposition is suggested by the lower frequency of the smallest fraction of artifacts (<2 cm) compared with experimental studies (SI Appendix, Fig. S5B). Several faunal specimens (25.2% of the assemblage) exhibit substantial smoothing and rounding, suggesting that some of the fauna are allochthonous. The stratigraphically discrete nature of the archaeological horizon is further supported by a high correlation between the variation in artifact elevation and that of the undulating elevation of the underlying clay horizon (r² = 0.89;n = 1424; SE: 4.49).

The age of the site is constrained by radioisotopic and magnetostratigraphic dating (Fig. 3 andSI Appendix, Fig. S3). The BD 1 archaeological horizon is 8.8 m stratigraphically above a 5–10 cm thick water-lain vitric tephra layer—the Ali Toyta Tuff—(SI Appendix, Table S2 and extendeddatasets S1 andS2) which yielded a⁴⁰Ar/³⁹Ar age on K-feldspar phenocrysts of 2.584 ± 0.034 Ma (95% confidence estimate) (SI Appendix, Figs. S7–S9 and extendeddataset S3). Paleomagnetic analyses of multiple sections at Ali Toyta from stratigraphically below the Ali Toyta Tuff to above the artifact horizon are all of normal polarity (Fig. 3 andSI Appendix, Figs. S9 and S10 and Table S3), which represent the upper Gauss Chron [C2An.1n, 3.032–2.581 Ma (11)]. Above this Chron, the reverse polarity Matuyama Chron has yet to be identified in the Ali Toyta drainage and may not be preserved. Therefore, the BD 1 assemblage was most likely deposited between the Ali Toyta Tuff (∼2.61–2.58 Ma) and the Gauss-Matuyama reversal (2.581 Ma). We conservatively adopt an age of ∼2.58 Ma for the assemblage noting, however, that it must be older than the Gauss-Matuyama reversal. This demonstrates that the BD 1 locality is older than previously documented Oldowan sites, all of which occur in the reverse polarity Matuyama Chron (2); (SI Appendix, Fig. S9). The age interval encompassed by these strata overlaps that of other sedimentary packages in the eastern LGRP region, ranging from 2.99 to 2.45 Ma (9,12).

The BD 1 faunal assemblage shares taxonomic similarities with the fauna from directly below the Gauss-Matuyama reversal in the Lee Adoyta fault block ∼5 km to the south (13). These assemblages indicate the proximity of open grassland in association with nearby lacustrine habitats (9) (SI Appendix, Table S4). This is consistent with previous studies that identified distinct environmental shifts relative to the earlier Hadar Formation (9). Only about half of the faunal specimens (53.4%) had 50% or more of the bone surface visible. This limited visibility makes it difficult to extract behavioral implications from bone surface modifications at BD 1 (SI Appendix, Fig. S11).

Artifacts from BD 1 show clear signs of conchoidal fracture including prominent bulbs of percussion, points of percussion, and contiguous flake scars on cores and on the dorsal surfaces of flakes (Fig. 4 andSI Appendix, Figs. S13 and S14 and Table S6). The majority of stone artifacts (64%) are made from a single type of fine-grained rhyolite that fractured predictably and has few internal flaws. Based on the occurrence of rhyolite pebbles and cobbles in conglomerate beds laterally continuous with the excavation horizon at BD 1, this material was readily available on the ancient landscape. Our analysis of these clasts (sampled up to 500 m from BD 1) shows that similar rock types are indeed present, but at lower frequencies, compared with their occurrence as artifacts in the BD 1 assemblage (SI Appendix, Figs. S3 and S12). The selection of specific rock types for artifact manufacture at BD 1 is also exhibited in later archaeological sites ranging from 2.5 to 1.78 Ma (14–16).

The BD 1 artifact assemblage is composed primarily of simple cores and flakes with smaller detached pieces (52%; <3 cm) being most abundant. The dominant flaking strategy was relatively simple with 45% of cores exhibiting flake scars on a single removal surface (SI Appendix, Fig. S14). Many flake scars end in step fractures (31%;SI Appendix, Fig. S15), which is consistent with many other Oldowan assemblages. Although flake scars on cores show clear indications of conchoidal fracture, most cores (62%) have only one or two flake scars (SI Appendix, Fig. S14). We identified no refits in this assemblage. Few artifacts show evidence of percussive activities that is not related to flaked stone tool manufacture (e.g., hammering or battering), and cores with multiple (>2) adjacent removals indicate some control of flaking (SI Appendix, Fig. S14). Core and flake sizes at BD 1 are similar to those found at younger Oldowan sites (SI Appendix, Fig. S16 and extendeddataset S4). Quantitative technological analyses of the BD 1 artifacts indicate that hominins had not yet mastered the skills that experimental studies have shown to be critical in the systematic production of sharp edges [e.g., location of percussion, identifying optimal platforms;Fig. 5 (17–19)]. However, in many respects, the overall technological pattern of the BD 1 assemblage most closely aligns with that of the earliest Oldowan sites (Fig. 6 andSI Appendix, Figs. S15–S17 and Table S5). In particular, when the next oldest Oldowan site (OGS 7), BD 1, and LOM3 are compared along the attributes that show the most variability in the early Oldowan (>2.0 Ma), BD 1 is more similar to the oldest Oldowan assemblages and Lomekwi falls outside the range of variability for this time period (SI Appendix, Fig. S18) (20).

The older 3.3 Ma LOM3 assemblage represents a distinctive technology compared with BD 1 and other Oldowan sites (Fig. 6). LOM3 includes extraordinarily large cores with relatively few flaking surfaces and extensive evidence of battering associated with percussive activities (7) (SI Appendix, Fig. S15). Many Oldowan sites, including BD 1, have a higher frequency of small flakes with discrete platforms (SI Appendix, Fig. S15) and limited evidence of percussive actions (Fig. 6). These features indicate that Oldowan technology represents a distinct technological pattern. Whether this is evidence for an independent invention of the Oldowan will depend on whether additional sites showing directional change can be found in the still rather substantial interval between LOM3 and BD 1. For now, though, these data combined with the recent documentation of tool use among a variety of primate species (5,6,21), suggest an initially diverse and potentially convergent set of tool-assisted behavioral adaptations among early hominins (22).

Discussion

The substantial assemblage of flaked stone artifacts from BD 1 marks the onset of hominin understanding of sequential flake removal and systematic flake production that is a characteristic of the Oldowan (Fig. 4 andSI Appendix, Figs. S13–S17 and Tables S7 and S8). In this regard, it is distinct from the much earlier LOM3 assemblage. LOM3, along with the contentious butchered bones from Dikika, Ethiopia (8), suggest that generalized tool use may be a shared ancestral trait of many hominins that is already present in the last common ancestor withPan. This may be reflective of a more generalized primate pattern of tool use (7,8,23). Although the earliest stone tool use may have enhanced the extractive foraging abilities of members of the genusAustralopithecus (24,25), as it does in many living primates (26–28), the ability to systematically produce sharp-edged flakes at BD 1 constitutes a derived trait likely related to new foraging strategies (29). Systematic flake production would have dramatically increased the foraging return for resources that can be processed with sharp-edged tools (1,30). This shift to a broader dietary adaptation may have resulted in a subsequent release of selective pressures on the earliest members of the genusHomo (31–33) and occurs concurrently with environmental shifts in northern Ethiopia (9,13).

BD 1 is the oldest known Oldowan assemblage. Although it is a single assemblage, further discoveries are likely to expose greater diversity in Pliocene tool use. Despite its antiquity, however, the BD 1 assemblage, along with the early Oldowan, is technologically distinct from older Pliocene technology and from the tool use seen in modern nonhuman primates. The link between Oldowan technology and preceding technologies, or to any other potentially ancestral primate tool use strategy (1,34), is currently unclear. Technological adaptations that are the hallmark of our genus, especially in the later Pleistocene (35), may be a derived feature with an independent origin relative to a series of diverse tool-assisted behaviors that began much earlier in the Pliocene (7,8).

Methods

Supplementary Material

Acknowledgments

We thank the Authority for Research and Conservation of Cultural Heritage, Ethiopian Ministry of Culture and Tourism, for permission to conduct fieldwork at Ledi-Geraru and laboratory research in the National Museum of Ethiopia, Addis Ababa, where the BD 1 artifacts are housed. This research was sponsored by National Science Foundation Grants BCS-1460502, BCS-1157346, BCS-1460493, and BCS-1157351. This paper was made possible through the support of a grant from the John Templeton Foundation. The opinions expressed in this paper are those of the authors and do not necessarily reflect the views of the John Templeton Foundation.

Footnotes

References

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.