The Acheulean is a temporally cohesive tradition

ABSTRACT

The Acheulean has long been considered a single, unified tradition. Decades of morphometric and technological evidence supports such an understanding by demonstrating that a single fundamental Bauplan was followed for more than 1.6 million years. What remains unknown is whether sites assigned to the Acheulean represent multiple socially-independent iterations of the same technological solution to shared ecological (functional) and ergonomic demands. Here, using the ‘surprise test’, the temporal cohesion of the Acheulean record is statistically assessed for the first time. Chronological data from 81 early and late Acheulean sites are investigated to see if breaks in this record warrant the designation of separate, culturally distinct groupings of sites. No significant results were returned, suggesting the Acheulean to be temporally cohesive and there to be no evidence of cultural convergence from a temporal perspective. When combined with previous morphometric, technological and spatial evidence, the best-fit scenario for the Acheulean continues to be that it represents a single, but variable, tradition.

KEYWORDS:

• Early Stone Age
• Lower Palaeolithic
• chronological exceptionality
• Gumbel domain of attraction
• cultural convergence
• convergent evolution

1. Introduction

The Acheulean has long been considered a single, unified cultural tradition that lasted for upwards of one and a half million years (Leakey 1971; Gowlett 1979, 2011; Isaac 1984). This understanding is dependent on the cultural information responsible for handaxe and cleaver technologies (i.e. bifacially flaked large cutting tools [LCTs]) having been continuously maintained by social learning processes from ~ 1.8 million years ago (Ma) through to ~ 0.2 Ma (Lycett et al. 2015; Lycett 2019; Shipton 2019; McNabb 2020). Lycett and Gowlett’s (2008) influential study of handaxe shape variation has been key to this idea. They demonstrated that although Acheulean handaxe forms do vary in time and space, little evidence for a fundamental divergence from an essential Bauplan exists. This finding built on earlier work by Isaac, Clark, and Gowlett (Isaac 1977; Crompton and Gowlett 1993; Clark 1994; Gowlett 2006) and has been supported on multiple occasions since (Wang et al. 2012; Gowlett 2015; McNabb and Cole 2015; Lycett et al. 2016; Hosfield, Cole, and McNabb 2018; Wynn and Gowlett 2018; García-Medrano et al., 2019; Key 2019; Shipton 2020; Shipton and White 2020; Caruana and Lotter 2022; although see: Corbey, Vaesen, and Collard (2016) and Gallotti (2016) for alternative perspectives). This includes evidence of the necessity of social learning processes for the efficient and effective reproduction of bifacial LCTs (Morgan et al. 2015; Schillinger, Mesoudi, and Lycett 2015; Lycett 2019; Pargeter, Khreisheh, and Stout 2019; Shipton 2019).

Evidence of a morphometric and technological Bauplan within the Acheulean is, however, only part of the theoretical foundation required to demonstrate a single cultural tradition existed. What remains unknown is whether sites (and artefacts) assigned to the Acheulean represent a single socially-maintained tradition, or whether multiple socially-independent iterations of the same technological solution are displayed in ‘the Acheulean’ archaeological record. Simply, it is potentially the case that multiple distinct lithic traditions (i.e. not linked by the social transmission of information) independently struck upon the use of bifacial LCTs during the Early-to-Middle Pleistocene, with morphological and technological similarities being driven by shared ecological (functional) and ergonomic demands (Wynn and Gowlett 2018). The invention and subsequent independent reinvention of Palaeolithic cultural traits is a frequently discussed phenomenon and forms part of a now long-lived discussion concerning cultural convergence and divergence in lithic technologies (see: Lycett (2015), O’Brien, Buchanan, and Eren (2018a), Groucutt (2020) and references therein); including the Acheulean (Petraglia and Shipton 2008; Lycett 2009a, 2010; Lycett and Norton 2010; Brumm and Rainey 2011; Wang et al. 2012; Blasco et al. 2013; Lieberman 2018; Arroyo, Proffitt, and Key 2019; Sharon 2019; Yoo 2019; Shipton 2020; White 2022). Reliably identifying instances of cultural convergence in the archaeological record is, however, challenging (O’Brien, Buchanan, and Eren 2018b).

Following the ‘out of Africa’ Acheulean dispersal model Shipton (2020) demonstrated that, consistent with other commentaries on the dispersal timings of the Acheulean (Clark 1994; Goren-Inbar et al. 2000; Lycett 2010; Shea 2010; Vallverdú et al. 2014; Kuhn 2021; White 2022), there is a negative relationship between the age of early Acheulean sites and their distance from east Africa. This built on earlier cladistic analyses by Lycett (2009b, 2010), who demonstrated there to be phylogeographic patterning in handaxe morphologies indicative of the Acheulean’s dispersal (social transmission) from east Africa. Both support the inference that the Acheulean represents a single tradition spread through social transmission from an east African source population. Further, Lycett and von Cramon-Taubadel (2008) demonstrate 45–50% of within-assemblage handaxe shape variation to be explained by geographic distance from east Africa. A fact consistent with an extended chain of social-transmission, and in turn, an absence of cultural convergence. These represent some of the few attempts to use spatial data to investigate the unity of the Acheulean.

One of the few unexplored routes to understanding whether the Acheulean represents a single cultural tradition is through temporal data. Specifically, we have no understanding of whether there are chronological gaps in the Acheulean record that warrant the designation of separate, socially independent groupings of sites. The logic being that if we see a significant gap in the temporal record of an artefact type (i.e. artefacts with shared shape and technological attributes), then the later instances could represent an episode of independent reinvention (cultural convergence) and not evidence of the maintenance of a single tradition through a continuous line of social learning. Indeed, all else being equal (e.g. search effort, preservation rates), the bigger the temporal gap in the archaeological record, the more likely it represents a true absence of cultural information.

It is entirely possible, therefore, that cultural convergence is responsible for the presence of multiple bifacial LCT-focused lithic traditions in the Pleistocene, and evidence of these distinct traditions may be present in the form of temporal gaps within the known archaeological record. Alternatively, if the Acheulean record can be demonstrated to be temporally cohesive, then it strengthens the argument for a single uninterrupted Acheulean tradition being present across its c. 1.6 million-year range.

1.1 How do you investigate temporal cohesion in a Palaeolithic tradition?

Cohesion and disunity in temporal data have been investigated from diverse perspectives. Of particular relevance here is the ‘surprise test’, a frequentist statistical method which assesses the temporal exceptionality of an occurrence relative to an alternative, often prior known, population or sample of temporal data. First introduced to Palaeolithic archaeology by Solow and Smith (2005), and more recently to palaeoanthropology by Roberts et al. (in press), the method most often tests the null hypothesis that a new record was generated by the same process that created previous records (Solow and Smith 2005). The method uses range and spacing data in a sample of consecutive temporal occurrences to identify whether an alternative earlier or later sample can be considered ‘surprising’ or ‘exceptional’ relative to the main sample (Figure 1).

Figure 1. A demonstration of how the surprise test is influenced by the temporal spacing and range of the archaeological sites forming the sample population against which the exceptionality of an alternative site is assessed. Figures 1(a,b) demonstrate the same small temporal gap, but due to the more constricted spacing and range of the early sites in 1B, this is the only record to produce a significant P-value. This is because the investigated gap is relatively greater in this scenario. The same is illustrated in Figures 1(c,d) under slightly different temporal circumstances and in the reverse temporal direction.

In the present scenario, the method can be used to suggest whether a bifacial LCT occurrence (site) was produced by the same cultural process responsible for a series of earlier or later occurrences. For example, it can test whether the 1.76 Ma Kokiselei 4 bifaces (Lepre et al. 2011) were likely to have been produced by the same cultural process responsible for later bifaces found at Konso, Olduvai Gorge and elsewhere (Beyene et al. 2013; Diez-Martín et al. 2015 [Table 1]). The test assumes sample occurrences to come from a distribution from the Gumbel domain of attraction, meaning the larger sample, against which an outlier is tested, is assumed by the test to represent the start or end of a cultural tradition (Figure 2).

Figure 2. A demonstration of how the temporal spacing and range of a sample population (site sample) influences the surprise test’s results, and in turn, the designation of cultural convergence. In Scenario 2A, two ‘outlier’ sites are investigated relative to an earlier sample, but only ‘Outlier 2’ is sufficiently separate from the sample population for it to support the inference that the original cultural tradition ended (as depicted by the modelled Gumbel distribution). Note that ‘Outlier 1’ would not exist in the tested scenario for ‘Outlier 2’. Scenarios 2B and 2C depict the same but in the reverse temporal direction.

Table 1. Archaeological sites and dates used in the early Acheulean analyses. The Acheulean occurrences are represented here by the ten earliest sites currently known. In addition, the 1.98 Ma age modelled by Key, Roberts, and Jarić (2021c) is also included; this represents the date beyond which the probability of the Acheulean occurring (based on current evidence) is <5%. In addition to the explanation outlined by Key, Roberts, and Jarić (2021c), these data do not include the recently dated ~1.6 Ma Acheulean artefacts from Melka Kunture (Gombore IB) (Mussi et al. 2021), which were excluded as they are ~100 m from the similarly aged Garba IV (Gallotti and Mussi 2018) occurrence that is included. A 1.67 Ma age has also been suggested for Acheulean artefacts at Oued Boucherit (Algeria) (Duval et al. 2021), however, this occurrence was excluded as its age relies on average sediment accumulation rates. Newly published dates from the T69 complex at Olduvai Gorge are not included due to their close proximity and temporal overlap with EF-HR (Fujioka et al. 2022).

Acheulean Site	Central Date	Date range	Reference
Modelled 5% Probability	1,980,000	2,080,000–1,880,000	Key, Roberts, and Jarić (2021c)
Kokiselei 4, West Turkana	1,760,000	1,810,000–1,720,000	Lepre et al. (2011)
KGA6-A1, Konso	1,740,000	1,740,000–1,660,000	Beyene et al. (2013)
FLK West, Olduvai Gorge	1,681,000	1,698,000–1,664,000	Diez-Martín et al. (2015); Stanistreet et al. (2018)
KGA4-A2, Konso	1,630,000	1,630,000–1,580,000	Beyene et al. (2013)
Level D, Garba IV	1,574,000	1,719,000–1,429,000	Gallotti and Mussi (2018)
Val River, South Africa	1,570,000	1,800,000–1,350,000	Gibbon et al. (2009)
OGS 12, Gona	1,550,000	1,600,000–1,500,000	Quade et al. (2004); Semaw et al. (2018)
MW2-L3, Melka Wakena	1,533,500	1,622,000–1,445,000	Hovers et al. (2021)
EF-HR, Olduvai Gorge	1,495,000	1,660,000–1,330,000	McHenry and Stanistreet (2018)
KGA10-A11, Konso	1,450,000	1,450,000–1,430,000	Beyene et al. (2013)

Cultural traditions often end when few individuals within a population retain the relevant information, and in turn, social-transmission ceases (Shennan 2015). This effectively means that the cessation of the Acheulean, and therefore the appearance of the aforementioned hypothesised gaps in the bifacial LCT record, are more likely to occur when the populations retaining this information are low in number. This means our best chance of identifying a culturally-significant temporal gap in the Acheulean record is soon after bifacial LCTs first emerge, or as their representation starts to decline in-line with the increased prevalence of prepared core technologies from c. 350 Ka (Figure 3). This means that studies investigating cultural convergence in the Acheulean should focus on its emergence and disappearance.

Figure 3. Competing cultural scenarios that explain the temporal record of bifacial LCT dominated Pleistocene sites. Figure 3(a) presents a series of bifacial LCT archaeological sites through time in two regions. Within Region 2, there is a gap in the archaeological record, potentially hinting at a loss of cultural information. This gap becomes the focus of investigation. Figure 3(b) depicts a single Acheulean tradition, with the occasional flow of cultural information between regions. Investigation of the temporal gap using the surprise test returned a P-value greater than α, suggesting there to be no break in the presence of Acheulean cultural information in Region 2. If the test returned a significant result, and cessation of the Acheulean was inferred in Region 2, it is possible that homologous bifacial LCT cultural information (i.e. the Acheulean) was transmitted back into the area via the flow of cultural information from Region 1 (Figure 3(c)). Figures 3(d,e) demonstrate scenarios where the test suggests the Acheulean to end in Region 2 but barriers to the reintroduction of homologous cultural information from Region 1 exist, meaning the later bifacial LCT sites in Region 2 could be the result of cultural convergence.

Moreover, for the majority of its existence the Acheulean is considered a global phenomenon (Wynn and Gowlett 2018), and if localised cultural extinctions did occur (e.g. Lycett and Norton 2010; Dennell 2018; Ashton and Davis 2021), then the later reappearance of bifacial LCTs in the same location could be due to the reintroduction of the same cultural information from elsewhere (Figure 3). While this later reappearance could also be due to convergence (i.e. analogy), there is no easy way to distinguish this from the above divergent (i.e. homologous) scenario using temporal data (O’Brien, Buchanan, and Eren 2018b). Thus, if we are to answer the question ‘does the Acheulean represent a single, temporally cohesive tradition’, we must focus on its emergence in east Africa and it’s regionally-disconnected ends.

The surprise test can be applied in both forward and reverse temporal directions and is free of scale limitations (Solow and Smith 2005), meaning it can be applied to any archaeological context. Here, the surprise test is applied to early and late Acheulean occurrences to investigate whether gaps in their temporal record provide evidence of cohesion or disunity.

2. Methods

In the present scenario the surprise test effectively asks, ‘how temporally surprising is an Acheulean occurrence (site) relative to the records observed before or after it’? Thus, and as recently explained by Roberts et al. (in press), the surprise test requires temporal information from all occurrences under investigation. This meant it was necessary to review previous literature concerning the earliest and latest known Acheulean occurrences. Importantly, this study is only concerned with the temporal cohesion of the Acheulean as defined for the majority of its existence (de la Torre 2016; Ollé et al. 2016), and thus, late Acheulean sites displaying additional technologies (e.g. Levallois) or transitional cultures were not included (e.g. Porat et al. 2010; Zaidner and Weinstein-Evron 2016; Sánchez-Yustos 2021; Duke, Feibel, and Harmand 2021; Herzlinger et al. 2021).

Temporal data from the earliest and latest known Acheulean sites were recently published by Key, Jarić, and Roberts (2021b, 2021c) and their samples form the foundation of the present analyses (Figure 2). These data include central age estimates for each Acheulean occurrence, along with upper and lower range limits as defined by the dating methods used for each site. Here, however, a distinct phenomenon is investigated and some sampling choices applied by Key, Jarić, and Roberts (2021b) are not appropriate to use. Thus, while the lower and upper date thresholds used by Key, Jarić, and Roberts (2021b, 2021c) are re-applied here to define the early and late periods of the Acheulean (respectively), the site number limitations for each country enacted by Key, Jarić, and Roberts (2021b) are not used here. Further, since the Key, Jarić, and Roberts (2021b, 2021c) studies came out in mid-2021 several additional early and late Acheulean occurrences have been published and the data used here have been updated accordingly (Table 1; Supplementary Table S1). Only directly dated Acheulean occurrences are included. See Key, Roberts, and Jarić (2021c) for further details of how temporal data and Acheulean sites were identified.

It is important to note that some sites included here are subject to ongoing debate. It is not, however, the point of this study to assess and critique individual archaeological occurrences. Here, the state of current, widely accepted Acheulean knowledge as evidenced through peer-reviewed literature is investigated. That is, the study investigates the field’s collective understanding of the Acheulean, and as such trust must be placed in the peer-review process as a way of fairly defining the data included in the models. Thus, some individuals may disagree with specific sites included here, but it is important to stress that these views will vary between different researchers.

Finally, it is important to acknowledge that the model assumes the earlier or later site occurrences, against which the exceptionality of an individual occurrence is tested, do not to contain an episode of cultural convergence. For example, if a date of 100 ka is investigated against four sites dated to 130, 135, 147 and 152 ka, then there is the assumption that an episode of cultural convergence is not observed between any of these pre-130 ka temporal gaps.

2.2 Site sample: early Acheulean

The oldest known bifacial LCTs, and thus instances of the Acheulean (de la Torre 2011), are found at Kokiselei 4 (Kenya (Lepre et al. 2011)), KGA6-A1 Konso (Ethiopia (Beyene et al. 2013)), and FLK West Olduvai Gorge (Tanzania (Diez-Martín et al. 2015)), and date to circa 1.75 Ma (Table 1; Figure 2). The presence of three chronologically constrained ‘earliest Acheulean occurrences’ suggest that the true origin of the Acheulean is not far beyond these known archaeological sites. Indeed, recent optimal linear estimation models concluded that the Acheulean likely originated close to 1.82 Ma, but had a 5% probability (based on current evidence) of being present ~1.98 Ma (Key, Roberts, and Jarić 2021c). Melka Kunture (Ethiopia), Val River (South Africa), Gona (Ethiopia) and other Acheulean sites in east and southern Africa are also present prior to 1.45 Ma (Table 1). The KGA10-A11 occurrence at Konso provides the youngest temporal record used in the early Acheulean sample (Beyene et al. 2013).

Temporal gaps in the early Acheulean record of Africa are ≤ 59,000 years (Table 1). The temporal range of the ten oldest known Acheulean occurrences above the 1.45 Ma lower threshold observed in Key, Roberts, and Jarić (2021c) analyses cover 310,000 years of the archaeological record. Two possible cultural scenarios can explain the temporal spacing of the early Acheulean record. Either a single cultural tradition is demonstrated, or one or more gaps in the early ESA record signify episodes of the Acheulean tradition being lost and bifacial LCT-dependent cultural traditions being independently reinvented by later populations.

To investigate the impact that a new, substantially earlier Acheulean occurrence may have on the exceptionality results, a hypothetical 1.98 Ma dated Acheulean occurrence was also included following Key, Roberts, and Jarić (2021c). The date range for this occurrence was set at ± 100,000 years. This produced a temporal gap of 220,000 years with Kokiselei 4.

2.3 Site sample: late Acheulean

Recent modelling suggests that late Acheulean occurrences disappear from the archaeological record at different times in different regions (Key, Jarić, and Roberts 2021b). The populations producing these late Acheulean tools were potentially restricted to isolated geographic pockets which may have prevented cultural transmission across large geographic distances (e.g. Santonja et al. 2016; Scerri et al. 2018). Investigation into the temporal cohesion of the late Acheulean should take this into account, meaning that multiple geographically distinct scenarios required investigation. Following Key, Jarić, and Roberts (2021b) four broad continental regions (Europe, Asia, north Africa and the Near East, sub-Saharan Africa) form the focus of the present late Acheulean analyses (Figure 4). In most continental areas there is a relatively tight temporal clustering of late Acheulean occurrences (Supplementary Table 1). This is expected given increased site preservation rates and high search effort in some regions relative to the early Acheulean in Africa.

Figure 4. The archaeological sites used in the present study imposed onto the distribution of the Acheulean as we currently understand it. Solid dots represent sites used in the early Acheulean models, while dots with a hollow centre represent sites used in the late Acheulean models. Those that are half full represent sites contributing to both scenario types.

In Europe, the 13 latest Acheulean sites are separated by ≤ 27,000 years, and eight of the observed temporal gaps are ≤ 8,000 years. The temporal range of the youngest known Acheulean sites in this region cover 142,500 years and end with the 162 ka occurrence of Arbo in Spain (Méndez-Quintas et al. 2019). The upper date threshold for site selection was 303 ka, as defined by the site of Broom (UK) (Green and Hosfield 2013; Key, Jarić, and Roberts 2021b) (Supplementary Table 1).

North African and Near Eastern late Acheulean sites are more tightly constrained yet, with consecutive young Acheulean sites in this region separated by ≤ 11,500 years and ranging across only 40,300 years. The most recent is observed 182 ka at EDAR 135, Sudan (Michalec et al. 2021), while the upper date threshold was defined by the 223 ka Sai Island occurrence in Sudan (Van Peer et al. 2003). This results in a sample of eight sites for this region (Supplementary Table 1). Note that Azraq and Holon are described as late Acheulean assemblages – and are thus are not ascribed to a transitional tradition – but both contain a small percentage of flakes that could be inferred as displaying ‘Levallois-like’ scar patterns (Copeland, 1991, 4; Porat et al. 1999).

The youngest Asian occurrences are present in the Korean Peninsula and are the subject of long-standing debate concerning their inclusion as part of the Acheulean. In light of this, these very late Korean sites are investigated separately (see below). The main Asian scenario includes late Acheulean sites from all other regions and Korean sites > 200 ka (i.e. Asia, with the exclusion of the latest occurrences in the Korean Peninsula). The maximum temporal gap between the youngest consecutive late Acheulean sites is 36,500 years, spread across a temporal range of 146,500 years. The upper date threshold, following Key, Jarić, and Roberts (2021b), was 221.5 ka and defined by the site of Jangnamgyo in Korea. This resulted in 12 sites being investigated (Supplementary Table 1).

Late Acheulean records from Sub-Saharan Africa present an outlier in terms of the gaps observed between temporally consecutive records. Indeed, a gap of 111,000 years is observed on one occasion, while five others are ≥ 40,000 years. The youngest Acheulean sites in this region are present across a 438,000-year period, with the youngest being the 212 ka site of Mieso in Ethiopia (de la Torre et al. 2014). The upper date threshold was 650 ka, as defined by Olduvai Bed IV (Njau et al. 2020). This resulted in 12 sites being investigated (Supplementary Table 1).

In addition to the continental-level models, there is a long-standing question concerning whether a large number of MIS 3 to MIS 5 dated bifacial LCT sites in the Korean Peninsula represent an episode of later post-Acheulean cultural convergence, or are part of the wider Acheulean tradition represented elsewhere in Asia (Bae 2017; Dennell 2018; Yoo and Lee 2022). To address this question, an additional scenario focusing on these late Korean sites was investigated. Temporal information from reliably dated in-situ bifacial LCT occurrences on the peninsular were extracted from Yoo and Lee (2022) and Lee (2017). Since many sites provided multiple dated bifacial LCT layers, data were collected from the earliest dated-layer only, as in many instances later reworking of substrate cannot be ruled out. This ensured that each LCT occurrence included in the models can be considered independent. Lee (2017) occasionally provides multiple dates for an artefact layer; in these instances, OSL data were preferentially sampled. Note that unlike the continental-level models, not all of these sites are ascribed to the Acheulean, but are included here due to the presence of bifacial LCTs and their role in the Korean Acheulean debate. This scenario investigated the exceptionality of late potential-Acheulean sites within the Korean Peninsula (n = 25), as well as their exceptionality relative to the 12 late Acheulean sites observed elsewhere in Asia (Supplementary Table 2).

Two possible cultural scenarios can again explain the above-described temporal spacing of the late Acheulean. Either a single cultural tradition is demonstrated in all regions and occurrences, or, one or more temporal gaps represent episodes of the Acheulean being lost and a new bifacial LCT focused tradition being independently reinvented at a later date.

In both the early and late Acheulean samples, sites were not excluded if they displayed the same age as another occurrence. However, the test does not allow for multiple same-aged occurrences to be included, meaning that when this occurred (n = 5), the age of one site was increased by + 1 (Supplementary Tables 1 and 2). Examination of temporal exceptionality between these same-aged instances is, therefore, not reflective of a difference observed in the archaeological record. The use of multiple similarly-dated occurrences will increase the likelihood of a significant value being returned, particularly when k is low. Finally, the present datasets represent the result of two (2020, 2022) exhaustive searches of the literature for relevant archaeological sites. Nonetheless, it is possible that a small number of valid sites may have been excluded here. The likelihood of this impacting the results obtained here are low and would likely only occur if multiple sites were missed within a temporally limited range.

2.4 Applying the surprise test: early Acheulean models

Following Solow and Smith (2005), a minimum of four consecutive Acheulean occurrences were used as the sample population against which to investigate the exceptionality of an alternative record (Figure 1). When λ is large a small k has good explanatory power in the surprise test (Solow and Smith 2005). However, given the known power trade-off observed between large and small k values the test was also run using additional k values up to k = 9. This was conducted in variable ways for different ranked Acheulean occurrences dependent on the number of sites include in each sample, with the test run in both forward and backward temporal directions. k was never below four and never above nine.

In the early Acheulean scenario there are 10 archaeological occurrences, which meant investigation of the 1^st, 2^nd, 3^rd, 4^th and 5^th oldest Acheulean sites were conducted in the reverse temporal direction, and in turn, the sample population used to investigate the exceptionality of these early occurrences was created from the 2^nd to 10^th youngest sites in the ‘early Acheulean’ sample, dependent on k. For example, if k = 5, then the five youngest records after the site investigated would create the sample population. Investigation of the oldest Acheulean site used k = 4, 5, 6, 7, 8, and 9, the 2^nd oldest site used, k = 4, 5, 6, 7, and 8, and so on until the 5^th oldest site used k = 4 and 5 (See Table 2 for further clarification). The 10^th, 9^th, 8^th, 7^th and 6^th earliest occurrences were conducted in the forward temporal direction, meaning that the 1^st to 9^th earliest Acheulean sites created the sample populations used to investigate the exceptionality of later occurrences. The youngest Acheulean site used k = 4, 5, 6, 7, 8, and 9, while consecutively older occurrences used one fewer k value until the 6^th oldest Acheulean site used k = 4 and 5 (Table 2).

Table 2. Significance values for all early Acheulean modelled scenarios using Solow and Smith’s (2005) surprise test for the exceptionality of a record relative to other dated occurrences. Note the reverse (↑) or forward (↓) temporal direction of the models, as when combined with the displayed k value it indicates the site sample used to test the exceptionality of the individual site’s P-value.

Occurrence	P (Central Estimates)						P (Resampling)						Model Direction
	k = 4	k = 5	k = 6	k = 7	k = 8	k = 9	k = 4	k = 5	k = 6	k = 7	k = 8	k = 9
Modelled 5% Probability	0.283	0.336	0.266	0.257	0.248	0.297	0.307	0.350	0.386	0.372	0.360	0.388
Kokiselei 4, West Turkana	0.888	0.859	0.856	0.854	0.877	0.897	0.806	0.856	0.853	0.849	0.862	0.872
KGA6-A1, Konso	0.570	0.566	0.563	0.637	0.697		0.758	0.753	0.751	0.772	0.789
FLK West, Olduvai Gorge	0.512	0.519	0.625	0.699			0.841	0.847	0.871	0.877
KGA4-A2, Konso	0.351	0.531	0.638				1.071	1.097	1.077
Level D, Garba IV	0.952	0.967					1.470	1.350					↑
Val River, South Africa	0.976	0.973					1.047	1.054					↓
OGS 12, Gona	0.857	0.890	0.883				1.222	1.232	1.229
MW2-L3, Melka Wakena	0.840	0.885	0.912	0.909			1.326	1.2696	1.255	1.246
EF-HR, Olduvai Gorge	0.404	0.686	0.764	0.815	0.811		1.106	1.069	1.064	1.055	1.045
KGA10-A11, Konso	0.554	0.484	0.681	0.751	0.800	0.798	0.972	1.021	1.015	1.007	1.009	1.009

The modelled 1.98 Ma age was not included in the exceptionality tests of known archaeological sites, but the 1^st to 9^th oldest known sites were used to create the sample populations against which the 1.98 Ma date was tested (reverse direction; k = 4 to 9). See Table 1 and Supplementary Table 1 for all data included in these models.

Sample populations in the forward and reverse models are assumed to represent the last or first (respectively) sites of a tradition, with the tested outlier being the hypothesised start or end (respectively) of an alternative tradition (Figure 1). In other words, the null hypothesis is that the outliers are part of the tradition represented by the main sample populations (i.e. all occurrences were generated by the same cultural process) (Figure 2). Further, the basic assumption of the model is that all records (discovered and undiscovered) follow the same Gumbel distribution, and that known records (i.e. the discovered sample population) represent a random sample of the whole record. Therefore, any increased representation of known sites through time are assumed to represent increased levels of bifacial LCT production in the Pleistocene. Relatedly, there is the assumption that site preservation levels and search effort (site discovery rates) are broadly equal within each investigated sample population, irrespective of whether k equals four or nine.

2.5 Applying the Surprise test: late Acheulean models

The same process was used to investigate the temporal cohesion of the late Acheulean occurrences. Here, however, five separate sets of data were analysed. These mirror the continental-level analyses conducted by Key, Jarić, and Roberts (2021b), meaning that late Acheulean temporal cohesion was investigated separately for sub-Saharan Africa, north Africa and the Near East, Europe, and Asia. In all late Acheulean models, interest lies in investigating the most recent occurrences of the Acheulean, meaning that the youngest half of each scenario’s occurrences were investigated in the forward temporal direction, while the earliest half were investigated in the reverse direction (see Table 3 for further clarification).

Table 3. Significance values for all central estimate modelled scenarios in the late Acheulean analyses. Note the reverse (↑) or forward (↓) temporal direction of the models, as when combined with the displayed k value it indicates the site sample used to test the exceptionality of the individual site’s P-value. See supplementary table S2 for the matching resampling data.

	P (Central Estimates)						Model Direction
	k = 4	k = 5	k = 6	k = 7	k = 8	k = 9
Europe
Broom (UK)	0.840	0.825	0.786	0.777	0.861	0.842
Cagny-l’Epinette (France)	0.884	0.848	0.840	0.910	0.896	0.928
Cien Fanegas (Spain)	0.139	0.148	0.467	0.401	0.563	0.748
Orgnac 3 (France)	0.437	0.800	0.757	0.850	0.925	0.919
Valdocarros (Spain)	0849	0.804	0.885	0.945	0.940
Pinedo (Spain)	0.999	0.999	0.999	0.999
Galería-GIIb (Spain)	0.926	0.966	0.961				↑
Harnham (UK)	0.817	0.935	0.929	0.931			↓
Cuxton (UK)	0.181	0.322	0.651	0.631	0.643
Tera River (Spain)	0.999	0.999	0.999	0.999	0.999	0.999
Porto Maior (Spain)	0.495	0.391	0.426	0.491	0.702	0.692
Lazaret Cave (France)	0.372	0.305	0.226	0.234	0.271	0.472
Arbo (Spain)	0.915	0.931	0.920	0.905	0.903	0.908
North Africa and Near East
Sai Island (Sudan)	0.789	0.843	0.898	0.903
Azraq (C-Spring, Jordan)	0.582	0.731	0.742
Wadi Arid (Egypt)	0.865	0.868
Kudaro I (Georgia)	0.899						↑
E. D. Atbara River (Sudan)	0.844						↓
Holon (Israel)	0.681	0.673
Saffaqah (Saudi Arabia)	0.422	0.563	0.557
EDAR 135 (Sudan)	0.747	0.745	0.801	0.797
Sub-Saharan Africa
Olduvai Bed IV (Tanzania)	0.411	0.419	0.723	0.825	0.843	0.836
Olorgesailie (Kenya)	0.781	0.926	0.957	0.961	0.958	0.953
Victoria Falls (Zambia)	0.745	0.847	0.859	0.848	0.825	0.844
Cave of Hearths (South Africa)	0.999	0.999	0.999	0.999	0.999
Hugub Bed (Ethiopia)	0.953	0.945	0.932	0.941
Garba 1 (Ethiopia)	0.603	0.509	0.581				↑
Manyara Beds (Tanzania)	0.310	0.285	0.466				↓
Kalambo Falls (Zambia)	0.283	0.385	0.352	0.467
Duinefontein 2 (South Africa)	0.730	0.658	0.693	0.669	0.723
Isimila (Tanzania)	0.929	0.912	0.892	0.901	0.893	0.909
Lower Herto (Ethiopia)	0.999	.0999	0.999	0.999	0.999	0.999
Mieso (Ethiopia)	0.637	0.803	0.843	0.823	0.798	0.812
Asia
Jangnamgyo (Korea)	0.156	0.508	0.484	0.461	0.573	0.542
Dingcun (China)	0.844	0.828	0.815	0.870	0.856	0.842
Singi Talav (India)	0.939	0.931	0.954	0.947	0.941	0.962
Kaldevanhalli (India)	0.875	0.919	0.905	0.892	0.933
Chongqing (China)	0.484	0.408	0.349	0.578
Patpara (India)	0.811	0.774	0.891				↑
Jwalapuram Tank (India)	0.817	0.832	0.918				↓
Bamburi 1 (India)	0.898	0.878	0.887	0.941
Bhimbetka Shelter (India)	0.665	0.620	0.577	0.600	0.756
Nakjhar Khurd (India)	0.930	0.967	0.963	0.959	0.961	0.977
Houfang (China)	0.951	0.951	0.976	0.974	0.971	0.973
Danjiangkou R. R. (China)	0.473	0.437	0.443	0.652	0.634	0.611

In addition, the temporal cohesion of bifacial LCT occurrences from the Korean Peninsula was investigated. Due to the increased number of occurrences in this scenario k = 5 in all instances. The test was run in both the forward and backward temporal direction (although this was not possible for the five earliest and latest occurrences; see Table 4 and Supplementary Table 2 for further clarification).

Table 4. Significance values for all modelled scenarios in the Korean late Acheulean scenario. All sites are from the Korean Peninsula, unless otherwise indicated. Note the reverse (↑) or forward (↓) temporal direction of the models, as when combined with the displayed k value it indicates the site sample used to test the exceptionality of the individual site’s P-value.

	Forward Temporal Direction ↓		Reverse Temporal Direction ↑
Site	Central Estimates (k = 5)	Resampling (k = 5)	Central Estimates (k = 5)	Resampling (k = 5)
Jangnamgyo			0.478	0.543
Dingcun (China)*			0.801	0.963
Singi Talav (India)*			0.925	1.073
Kaldevanhalli (India)*			0.856	0.857
Chongqing (China)*			0.516	0.522
Chongokni (E55s20IV)	0.639	1.048	0.890	0.906
Patpara (India)*	0.892	0.925	0.759	0.828
Jwalapuram Tank (India)*	0.824	0.892	0.743	0.828
Bamburi 1 (India)*	0.892	0.916	0.133	0.433
Bhimbetka Shelter (India)*	0.629	0.633	0.875	1.397
Jangsanni	0.961	0.984	0.943	1.204
Nakjhar Khurd (India)*	0.966	1.101	0.948	0.985
Houfang (China)*	0.950	1.172	0.815	0.923
Jeongokri ACF	0.863	1.060	0.959	0.972
Galdun	0.840	1.305	0.548	0.661
Wolso	0.294	0.677	0.375	0.659
Danjiangkou R. R. (China)*	0.393	0.530	0.891	0.962
Sangjiseokri	0.952	1.040	0.478	0.688
Jeongokri Jung 2–5	0.722	0.906	0.926	1.115
Chongokni (R. 1–2, SL 4)	0.983	1.063	0.344	0.957
Chuwoli	0.838	0.958	0.835	1.048
Mansur	0.954	1.209	0.585	0.847
Segeori	0.740	0.950	0.986	1.486
Chongokni (R. 2–5, SL 4)	0.986	1.184	0.999	1.296
Unjeong (Loc. 1 SL 7)	0.999	1.271	0.943	1.150
Maaeri	0.889	1.494	0.492	0.967
Wondangri	0.429	0.680	0.824	1.298
Gawolri	0.625	0.831	0.760	0.810
Nosanri (Loc. 4 SL 6)	0.625	1.250	0.799	0.909
Namgyyeri	0.740	1.115	0.924	0.959
Kumpari	0.913	1.160	0.767	0.769
Baeki	0.678	1.089	0.661	0.673
Mansuri (Loc. 13 SL 2)	0.671	0.929
Jangnamgyo (Loc. 1. SL 6)	0.618	0.660
Local 26–5, UDSC Sites	0.651	0.700
Janggidong (Loc. 5)	0.897	0.897
Jeungsan	0.914	0.914

2.6 The surprise test

Solow and Smith (2005) and Roberts et al. (in press) detail the formulaic expression of the surprise test, which is replicated here, adjusted for the present scenario. The text below details the procedure used to test in the forward temporal direction. For tests in a reverse direction, chronologically relevant instruction should be altered accordingly.

Let t₁> t₂> … >t_k be the k most recent occurrences of the Acheulean ordered from the most recent to the earliest. Solow and Smith’s (2005) method assumes that these represent the k smallest values of a larger collection of values generated from a distribution from the Gumbel domain of attraction. The Gumbel distribution represents an extreme value distribution often used to model maximum and minimum values of a sample or population. Suppose that a more recent specimen is dated at time y, interest centers on assessing the exceptionality of this more recent specimen. Under the null hypothesis that the new case was generated by the same process as the earlier cases, Solow and Smith (2005) showed that the quantity,

$$S_k=\frac{y-t_1}{\left(y-t_1\right)+{\Sigma }_{j=1}^{k-1}\left(j+1\right){\left(t_j-t_{j+1}\right)}^{\text{'}}}$$

has a β distribution with parameters 1 and k-1 so that the P-value (α = 0.05) corresponding to an observed value S_k is

$P=\hspace{0.17em}{{\left(1-S_k\right)}^k}^{-1}.$

Due to most Acheulean occurrences being represented by a central age value derived from a broader range, a resampling approach was used in addition to the surprise test being applied to central age values. The resampling procedure required a date to be randomly drawn from a normal distribution bounded by a site’s age range, and this being repeated for each site included in the analysis. The use of a normal distribution assumes that the more extreme values in a date range less accurately represent the age of a site relative to more central values. The central age value was used as the mean, and the standard deviation was half of the difference between the central estimate and range bounds. The randomly generated dataset was assessed using the surprise test as per the standard procedure outlined above. This process was repeated 10,000 times and the results were expressed as a mean across all iterations. All analyses were undertaken in R version 4.1.2 (see: Roberts et al. in press for the associated code).

3. Results

3.1 The temporal exceptionality of early Acheulean sites

All modelled scenarios returned a P-value greater than α (i.e. 0.05) (Table 2), suggesting the early Acheulean record of Africa to be temporally cohesive. That is, there are no temporal gaps substantial enough to indicate one site as being surprising or exceptional relative to prior or later Acheulean occurrences. Thus, the temporal record of the early Acheulean does not currently demonstrate evidence of cultural convergence, and all sites between c. 1.4 and 1.8 Ma can be considered to belong to the same cultural process.

Tests using central estimate data were typically closer to significance than those performed through the resampling procedure (Table 2). The smallest P-values were often when k = 4, with KGA4-A2 at Konso (Ethiopia) and EF-HR at Olduvai Gorge (Tanzania) returning the lowest values of P = 0.351 and 0.404, respectively. Tests investigating Key, Roberts, and Jarić (2021c) modelled 1.98 Ma date returned the values closest to significance (P = 0.248–0.388), but were nonetheless non-significant, indicating that given the porosity of the current ESA archaeological record a temporal gap of 220,000 years is not enough to designate a site as surprising or unexpected.

3.2 The temporal exceptionality of late Acheulean sites

All modelled late Acheulean scenarios returned a P-value greater than α (i.e. 0.05) (Table 3; Supplementary Table 3), suggesting these late bifacial LCT dominated sites to be temporally cohesive within their respective regions. Once again, then, there are no chronological breaks of duration great enough – relative to the temporal spacing and range of sites in the comparative samples – to indicate any late Acheulean sites as being surprising or unexpected. In turn, suggesting that from c. 300 ka in Eurasia and 650 ka in sub-Saharan Africa, there is currently no temporal evidence to support the presence of cultural convergence in the Acheulean archaeological record.

Tests conducted using central estimate data were, again, typically closer to significance than those produced using the resampling data. Values did approach significance in a number of occasions, nearing 0.100 in three European and two Asian/Korean tests. Such values were only returned when k = 4 or 5. The sub-Saharan and north African and Near Eastern records appear more strongly cohesive with their lowest P-values equalling 0.283 and 0.422, respectively.

4. Discussion

Presented here is an investigation into the temporal cohesion of the Acheulean archaeological record. It was undertaken to identify whether any gaps in the current record warrant the designation of independent – and temporally differentiable – bifacial LCT focused cultural traditions. Identification of such a gap could, in turn, be attributed to a break in the social transmission of Acheulean cultural information during the Pleistocene, potentially suggesting the presence of cultural convergence during the bifacial LCT artefact record. In effect, the study tested whether archaeological sites currently assigned to the Acheulean should continue to be considered part of a single tradition. The study forms part of a larger movement to better integrate temporal modelling into our understanding of human behaviour and evolution during the Pleistocene (e.g. Du and Alemseged 2019; Bevan and Crema 2020; Faith et al. 2021; Key, Roberts, and Jarić 2021a; Bobe and Wood 2022; Crema 2022; Surovell et al. 2022; Djakovic, Key, and Soressi 2022).

4.1 Temporal cohesion in the early Acheulean

The early Acheulean archaeological record appears to be temporally cohesive between 1.75 and 1.4 million years ago. Indeed, the null hypothesis that all sites were produced by the same cultural process has not been rejected and no individual archaeological sites can be considered surprising or exceptional relative to samples of earlier or later Acheulean occurrences (Table 2). Thus, currently, there is no evidence of a break in the social transmission of Acheulean information during this period, and as far as temporal evidence can demonstrate, the early bifacial LCT record of Africa should continue to be considered part of a single cultural tradition.

Early Acheulean occurrences in Asia, Europe and the Near East have not been investigated here, but the emergence of bifacial LCTs in these regions after the Acheulean’s establishment in Africa makes the single-tradition hypothesis unfalsifiable using temporal data. If the surprise test did pick up a localised break in social transmission in Eurasia, it would be impossible to rule out the reintroduction of the same Acheulean cultural information from elsewhere (e.g. Villa 2001; Lycett and von Cramon-Taubadel 2008; Hosfield 2011; Dennell 2018; Ashton and Davis 2021). In this scenario (Figure 3(c)), the reappearance of bifacial LCT production in a region represents the introduction of homologous (i.e. of shared, common ancestry) cultural information (O’Brien, Darwent, and Lyman 2001; Lycett 2015; O’Brien, Buchanan, and Eren 2018b). This reasoning has been strengthened through the present analyses, which suggests the Acheulean to have been continuously present in Africa between 1.75 and 1.4 Ma. Post 1.4 Ma in all regions (but prior to c. 300 ka), including Africa, the same logic applies; it is not possible to use temporal data alone to assess cultural convergence as Acheulean cultural information could easily be reintroduced from elsewhere following a localised extinction event (Figure 3(c)). Future work may wish to use the surprise test to identify instances of regional Acheulean extinctions between 1.4 and 0.3 Ma, even if the presence of cultural convergence cannot be tested.

Given the sparse but regular, marine-isotope-stage-dependent, temporal spacing of early Acheulean sites in Europe (Mosquera et al. 2016; Ashton and Davis 2021), it is unlikely that the surprise test would have returned significant results had it been applied to this region. The test may potentially have returned significant results in Asia due to sporadic early occurrences such as Attirampakkam (Pappu et al. 2011) and Bose Basin (Wang and Bae 2015), but highly irregular sample populations would simultaneously result in substantial gaps being required before significant results are returned.

Models based on the current ESA record estimate the Acheulean to have originated 1.82 Ma (Key, Roberts, and Jarić 2021c), with known artefactual evidence from 1.75 Ma. As Key, Roberts, and Jarić (2021c) readily acknowledge, these modelled dates could be updated following future discoveries, but they are unlikely to move beyond the published confidence intervals. What the present analyses have demonstrated is that should a bifacial LCT site earlier than West Turkana (Lepre et al. 2011), Konso (Beyene et al. 2013), or Olduvai Gorge (Diez-Martín et al. 2015) be discovered then it should be considered part of the Acheulean tradition, even if it dates to as early as 1.98 Ma (assuming site ranges and spacing remain similar). That is, temporally there would be no reason to consider it to be an independent cultural entity. The fact that a 220,000-year gap is not enough to identify a site as temporally exceptional speaks to the high porosity of the current ESA record. Indeed, it is worth reemphasising that the present modelled distributions are derived from the known archaeological record. Should the resolution of the ESA record after 1.75 Ma increase, then such a large gap may in the future suggest the absence of cultural information.

4.2 Temporal cohesion in the late Acheulean

It is reasoned that towards the end of the Acheulean episodes of cultural convergence could again be revealed due to bifacial LCT focused populations becoming increasingly fragmented (Figure 3(d)). As Middle Stone Age and Middle Palaeolithic prepared core technologies dispersed through Africa and Eurasia after c. 350 ka (Deino et al. 2018; Adler et al., 2014; Akhilesh et al. 2018; Moncel et al. 2020), ‘purely Acheulean’ lithic practices would have become regionally isolated. In turn, when low population densities, localised extinctions, ecological change, or other demographic and environmental factors resulted in the regional loss of Acheulean cultural information (Lycett and von Cramon-Taubadel 2008; Dennell 2018; Lycett and Norton 2010; Kolodny, Creanza, and Feldman 2015; Shennan 2015; Derex and Mesoudi 2020), it may no longer have been possible for the same information to be reintroduced from another region (Figure 3(d,e)). If the surprise test revealed significant temporal gaps during this period (i.e. gaps indicating multiple temporally distinct cultural groupings), then convergent evolutionary processes may have been responsible for the later-dated biface artefacts.

No instances of temporal disunity were identified in any of the late Acheulean scenarios, and thus, no null hypotheses were rejected. That is, all individual late Acheulean occurrences were found to fit within expected temporal frameworks as defined by prior or later samples of sites. Once again, then, all sites appear to be produced by the same cultural processes and there is no temporal evidence for a break in the social transmission of Acheulean information. These analyses were conducted at continental-level regional scales and their temporal coverage varies, but broadly these results can be applied from c. 300 ka until the Acheulean ends in each region. The African analyses are the exception, with late Acheulean temporal cohesion being inferred from 650 ka onwards.

On several occasions P-values were closer to significance than those observed in the early Acheulean analyses. This included the temporal gap between Orgnac 3 in France (Michel et al. 2013) and Cien Fanegas in Spain (Moreno et al. 2016) (P = 0.139), and Harnham and Cuxton in the UK (Wenban-Smith, Bates, and Marshall 2007; Bates et al. 2014) (P = 0.181). While these instances do not strongly suggest a break in social transmission, their temporal gaps are relatively greater than those seen elsewhere, potentially hinting at a modest demographic or cultural signal. Demographically it could suggest low population levels, and in turn, low numbers of Acheulean bifaces being produced (and thus found) during this period. Alternatively, it could hint at a currently archaeologically invisible (or at least obscured) break in social transmission, that could be more strongly supported at a later date through new site discoveries creating greater date-densities before or after these breaks. Whatever the case, these instances do not currently support the suggestion of bifacial LCT-focused lithic cultures being independently reinvented in Europe.

An additional set of analyses focusing on well-dated but late-occurring bifacial LCT sites in the Korean Peninsula was undertaken. These sites have been subject to extensive discussion concerning their inclusion in the wider Acheulean (Norton et al., 2006; Bae 2017; Yoo and Lee 2022). Prior investigation has focused on their morphological and technological attributes, which identified some differences relative to more-securely attributed and more-westerly located Acheulean occurrences (Norton et al., 2006; Lycett and Gowlett 2008; Shipton and Petraglia 2011; Li et al. 2018). Given their still-considerable form overlap with Acheulean LCTs, data key to the Korean artefact’s non-Acheulean designation is their late < 100 ka age (Yoo 2019; Yoo and Lee 2022). Despite recent modelling supporting the persistence of the Acheulean in Asia after 100 ka (Key, Jarić, and Roberts 2021b).

The present analyses tested 1) whether these late Korean bifacial LCT examples can be considered temporally exceptional or distinct relative to the Acheulean record observed elsewhere in Asia, and 2) whether the Korean sample itself is temporally cohesive. No significant results were returned, and the combined Asian and Korean bifacial LCT records should – at least from a temporal perspective – be considered a single cultural tradition. That is, there are no gaps in the archaeological record that suggest the presence of an independent, culturally distinct grouping of Korean biface sites. Not only is there chronological overlap between the Korean, Indian and Chinese biface records (Supplementary Table 2), but when only Korean records are present (i.e. after 100 to 75 ka) there is still no evidence to suggest a break in social transmission. Thus, at present, the Asian archaeological record supports the continuous transfer of Acheulean cultural information from c. 200 to 30 ka and, in lieu of improved temporal records, questions concerning the distinctiveness of Korean (and other Asian) bifacial LCT sites should focus on technological, morphometric and spatial data (e.g. Lycett and Norton 2010; Shipton and Petraglia 2011; Wang et al. 2012; Lee 2017; Yoo 2019).

4.3 The Acheulean as a temporally cohesive cultural tradition: implications and limitations

Presented here is the current state-of-the-art understanding given the known archaeological record. It is important to highlight that future archaeological discoveries have the potential to alter the results and conclusions reported here. In India alone, there are more than 1500 Lower Palaeolithic sites lacking absolute dates (Chauhan 2020), with only a small number having been excavated. Should a series of these return c. 100 ka ages, then the surprise test may subsequently suggest the presence of temporally-distinct groupings of bifacial LCT sites in Asia. In turn, the Korean examples may yet be inferred to be the result of cultural convergence. Equally, however, this may not occur and the present scenario may continue to be supported through evenly distributed site discovery rates. There is, therefore, potential for instances of cultural convergence to be ‘hiding in plain sight’ in the current Acheulean record, but whether such instances are revealed or not remains to be seen.

The data presented here challenge previous suggestions that the sporadic and patchy Acheulean record of some regions indicates the presence of multiple bifacial LCT traditions or the absence of the Acheulean tradition during specific periods (e.g. Lycett and Norton 2010; Gallotti 2016; Moncel and Schreve 2016; Dennell 2018; Yoo and Lee 2022). What has instead been demonstrated is that because the archaeological record is at times sparse, the opposite should be accepted until such a time that a more complete temporal picture is available. Indeed, a temporal gap should only be inferred to be culturally meaningful when site distributions on either side indicate it to be exceptional or surprising. Thus, a temporally unified and cohesive Acheulean archaeological record should be the null hypothesis unless there is strong evidence from temporal or other sources to suggest otherwise (an example of which could be Palaeoclimatic data in northern Europe [Ashton and Davis 2021]).

Despite multiple late Acheulean scenarios being investigated here, other more spatially-refined scales of analysis (for example, Iberia or southern Africa) could suggest episodes of cultural loss, particularly if performed across greater date ranges. But as demonstrated in Figure 3, cultural convergence could only be supported if Acheulean information could not be reintroduced from another region. Further, small-scale regional Acheulean investigations can present their own additional limitations (See: Ollé et al. 2016).

The present results do not conclusively mean that there were no instances of bifacial LCT technologies being independently reinvented during the Pleistocene. They reveal that such instances are not currently evidenced using temporal data, and use of chronological information to suggest otherwise appears unsupported. When combined with morphological, technological, spatial and other information (Vaughan 2001; Gowlett 2006, 2015; Lycett and Gowlett 2008; Petraglia and Shipton 2008; Lycett 2010; de la Torre 2016; Finkel and Barkai 2018; Hosfield, Cole, and McNabb 2018; Wynn and Gowlett 2018; García-Medrano et al., 2019; Shipton 2020; White, 2022), the best-fit scenario for the Acheulean continues to be that it represents a single, but variable, cultural tradition ‘handed over from tool-maker to tool-maker over an exceptionally long period’ (Lycett and Gowlett 2008, 309).

5. Conclusion

Using a novel temporally-based evidential framework, this study supports the Acheulean’s designation as a single cultural tradition. It does so by demonstrating the Acheulean record to be temporally cohesive, with no currently-known early or late bifacial LCT sites displaying an exceptional or surprising temporal presence relative to other known sites. Simply, it was not possible to reject the null hypothesis that all investigated bifacial LCT occurrences were produced by a single cultural process. When combined with previous morphometric, technological and spatial assessments of the Acheulean, the best-fit scenario for explaining more than 1.6 million years of bifacial LCT production continues to be that a single, but variable (Isaac 1977; Vaughan 2001; Lycett and Gowlett 2008; Gowlett 2015), lineage of cultural information existed. However, as discussed above, the present results are open to both change or stasis as additional sites are discovered and dated in the future. Other archaeological phenomena could profitably employ the methods described here. As the surprise test is free of scale limitations, it can easily be applied to later Prehistoric or historic periods.

Acknowledgments

I am grateful to David Roberts and Ivan Jarić for introducing me to the surprise test. I would like to thank the anonymous reviewers for their feedback on an earlier version of this article.

Disclosure statement

No potential conflict of interest was reported by the author.

Data availability statement

All data are available in the text or supplementary information. The relevant code is available in Roberts et al. (in press).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/00438243.2023.2169340.

Additional information

Notes on contributors

Alastair Key

Alastair Key is an Assistant Professor of Palaeolithic Archaeology in the Department of Archaeology, University of Cambridge. He research investigates the behaviour and evolution of early humans through the excavation and analysis of lithic artefacts, the experimental production and use of replica stone and organic tools, and a variety of different modelling processes.