The transition from hunting–gathering to agriculture in Nubia: dental evidence for and against selection, population continuity and discontinuity

Abstract

Some researchers posit population continuity between Late Palaeolithic hunter–gatherers of the late Pleistocene and Holocene agriculturalists from Lower (northern) Nubia, in northeast Africa. Substantial craniodental differences in these time-successive groups are suggested to result from in situ evolution. Specifically, these populations are considered a model example for subsistence-related selection worldwide in the transition to agriculture. Others question continuity, with findings indicating that the largely homogeneous Holocene populations differ significantly from late Pleistocene Lower Nubians. If the latter are representative of the local populace, post-Pleistocene discontinuity is implied. So who was ancestral to the Holocene agriculturalists? Dental morphological analyses of 18 samples (1075 individuals), including one dated to the 12th millennium BCE from Al Khiday, near the Upper Nubian border, may provide an answer. It is the first Late Palaeolithic sample (n = 55) recovered within the region in approximately 50 years. Using the Arizona State University Dental Anthropology System to record traits and multivariate statistics to estimate biological affinities, Al Khiday is comparable to several Holocene samples, yet also highly divergent from contemporaneous Lower Nubians. Thus, population continuity is indicated after all, but with late Pleistocene Upper—rather than Lower Nubians as originally suggested—assuming dental traits are adequate proxies for ancient DNA.

1. Introduction

The hypothesis of population continuity in Nubia from the late Pleistocene through to the Holocene is greater than 50 years old. Arguments against it began over 40 years ago and the back-and-forth continues. This is remarkable, considering that to many outside observers the subject would seem esoteric and the location obscure. However, the issue at the heart of this debate is of interest to many, entailing in a microcosm a core event in world history—the transition from hunting–gathering to agriculture [1–3]. Specifically, are marked morphometric changes in skulls and teeth of Nubian skeletal remains before and after this shift a textbook example of regional diet-related selection, or population discontinuity?

Ancestry between Late Palaeolithic (20 000–12 000+ BP [4,5]) Nubians of the late Pleistocene and Meroitic (300 BCE–400 CE) through later populations was proposed by Greene [6,7] and others [8–10]. As a region, Nubia parallels the Nile River from the First Cataract at Aswan, Egypt, to the Blue and White Nile Rivers [11–13] near Khartoum, Sudan (between dashed lines in figure 1). It is then bisected into sub-regions, Lower (north) and Upper (south) Nubia [14]. The continuity hypothesis pertains generally to the former, particularly the area straddling the Egypt/Sudan border. It is from there that skeletal samples studied by continuity advocates were recovered, including Late Palaeolithic Wadi Halfa (3 km southwest of Gebel Sahaba in figure 1). Support for this hypothesis was provided in later studies [1,15–18]. Larsen [19, p. 312] agreed, stating that ‘studies of craniofacial morphology, biological change and population history’ is suggestive of continuity.

Figure 1.

Locations of place names relative to samples in table 1. (Online version in colour.)

Nonetheless, much of the evidence is based on dentitions, including ostensibly shared ‘rare cusp variants’ in Late Palaeolithic and Meroitic samples [7, p. 322]. The first hypothesis to explain the differences between hunter–gatherers and agriculturalists—from cranially robust with large morphologically complex teeth, to gracile with globular crania and smaller simpler dentitions—was proposed to result from dental-related selection. The former population was said to have experienced selective pressures favouring large teeth with incisor shovelling, extra cusps and other complex occlusal features to resist crown wear from dietary grit [6]. With the advent of agriculture and higher-carbohydrate foods, Holocene populations were then said to have undergone reduction, to smaller, simpler caries-resistant teeth. This ‘caries selection hypothesis' involved the molars, where fewer cusps and simpler groove patterns ([7, p. 323]; [17,20]) presumably provided reproductive benefits.

Next came the ‘masticatory-functional hypothesis’ ([8, p. 502]; [9,10]). It posited that dental reduction was a side effect of changes to the craniofacial complex after transitioning to agriculture. In order of importance, the change from hunting–gathering yielded: (i) reduction in masticatory muscle size, (ii) reduction in jaw growth for a less projecting midface, (iii) secondary changes to more globular crania, and (iv) smaller teeth and jaws, which decreased prognathism [8,9].

Focus then returned to teeth ([1, p. 513]; [18)], which were said to be ‘under selective pressure to reduce in size to avoid the negative sequelae of dental crowding’, not caries. Explicitly, all teeth decreased in size. Then during the 5000-year progression from agriculture to intensive agriculture, only molars reduced in support of caries selection [1]. Others [16] agree concerning dental reduction, though in response to general stress among agriculturalists based on more developmental pathologies than hunter–gatherers. Beyond conference papers, research supporting Nubian continuity then lessened—until recently. In 2016, Galland et al. [3] revisited the topic, finding cranial geometric morphometric analyses in five samples, including Wadi Halfa, again support the masticatory-functional hypothesis (but see below).

Conversely, while it is known craniodental reduction occurred worldwide in the agricultural transition [2], the amount of change in Lower Nubians since the Pleistocene is striking. Armelagos et al. [17, p. 1] noted ‘the reduction in facial morphology…is greater by several orders in magnitude than the reduction in general body size’. Such major change prompted Hillson [21, p. 91] to remark ‘although [in situ evolution] is an attractive idea, there is doubt as to whether there would be an adequate number of generations to achieve it’. Later dental [22–27], cranial [28,29] and post-cranial [30] research supported this statement, finding Late Palaeolithic Lower Nubians too divergent from later Holocene peoples to be ancestral.

It should then follow that population discontinuity occurred after the Pleistocene [22–26]. In the yet-largest regional analyses of 12 Lower and Upper Nubian samples, including one from Late Palaeolithic Gebel Sahaba (n = 49 individuals) and Tushka (n = 18) (figure 1), such an occurrence took place in the Neolithic, before 4600–4000 BCE (i.e. Final Neolithic (table 1) according to [31], or Middle Neolithic in [32]); homogeneity is then evident through the Christian period (CE 550–1350) [33]. However, as a contradictory hypothesis, discontinuity would assume Wadi Halfa and Gebel Sahaba/Tushka peoples were indigenous to the region to have had the potential to contribute genetically to subsequent peoples. A-Group and later agriculturalists would then have represented ‘an arrival of new people at the advent of the shift from hunting–gathering to farming’, in the ‘population influx’ hypothesis of Galland et al. [3, p. 7]; like the masticatory-functional hypothesis, the authors relate that it also aligns with their analyses. However, most later studies suggest the population represented by Late Palaeolithic (Lower) Nubians originated extra-regionally [22–26,28–30,33]. As above, research along these lines then waned—until recently [34]. Larsen [35] also amended his earlier remarks [19], acknowledging the potential for population discontinuity.

Table 1.

Dental samples from Nubia used in the present study.

sample	site(s)/region of origin	cultural affiliation	subsistence	date	no.	curation
Lower Nubia
Gebel Sahaba (GSA)a	Gebel Sahaba (sites 117, 8905) and Tushka	Late Palaeolithic	hunter–gatherers and fishing	>11 600 BPb	67	BMc
Gebel Ramlah (GRM)	Gebel Ramlah	Neolithic	intensive collectors/ pastoralists	4600–4400 BCE	82	GRM
A-Group (AGR)	Faras to Gamai	A-Group	agro-pastoralists	3800–2900 BCE	52	PAN
C-Group (CGR)	Faras to Gamai	C-Group	agriculturalists	2300–1580 BCE	62	PAN
C-Group (HCG)	Hierakonpolis	C-Group	agriculturalists	2300–1580 BCE	56	HK
Pharaonic (PHA)	Faras to Gamai	Pharaonic	agriculturalists	1550–1070 BCE	38	PAN
Meroitic (MER)	Semna, Faras to Gamai	Meroitic	intensive agriculturalists	100 BCE–350 CE	94	ASU, PAN
X-Group (XGR)	Semna, Faras to Gamai	X-Group	intensive agriculturalists	350–550 CE	63	ASU, PAN
Christian (CHR)	Semna, Faras to Gamai	Christian	intensive agriculturalists	550–1350 CE	41	ASU, PAN
Upper Nubia
Al Khiday (AKH)	Al Khiday	Late Palaeolithic	hunter–gatherers and fishing	12th millennium BCEb	55	UP
Al Khiday (AKN)	Al Khiday	Neolithic	agro-pastoralists	5000–4300 BCE	28	UP
El Ghaba (GHB)	Near Khartoum	Neolithic	agro-pastoralists	5600–4300 BCE	119	LJMU
R12 (R12)	Site R12 at Kawa	Neolithic	agro-pastoralists	5300–4300 BCE	50	BM
Kerma (KAM)	Near Kawa	Kerma Ancien/Moyen	agriculturalists	2500–1750 BCE	60	BM
Kerma (KMC)	Kerma	Kerma Classique	agriculturalists	1750–1500 BCE	63	CAM
Soleb (SOL)	Soleb	Pharaonic	agriculturalists	1550–1070 BCE	32	MH
Tombos (TOM)	Tombos, 3rd Cataract	Napatan	agriculturalists	850–664 BCE	50	PUR
Kushite (KUS)	Kawa, Gabati	Meroitic/post-Meroitic	intensive agriculturalists	600 BCE–550 CE	63	BM

^aSample three-letter abbreviations used in the electronic supplementary material, tables and figures.

^bDates in BP or millennia BCE used owing to difficulties in obtaining absolute dates for these samples (see text for details).

^cASU, Arizona State University; BM, British Museum; CAM, Cambridge University; GRM, Gebel Ramlah site, Egypt; HK, Hierakonpolis archaeological site; LJMU, Liverpool John Moores University; MH, Musée de l’Homme; PAN, Panum Institute; PUR, Purdue University; UP, University of Padova.

It thus seems that additional analyses are needed to move this debate along. However, new evidence is critical. In this study, the first ‘new’ Late Palaeolithic regional sample in approximately 50 years, from Al Khiday (figure 1), is analysed. Using 36 nonmetric traits from the Arizona State University Dental Anthropology System (ASUDAS) [36], these remains are compared with those from Gebel Sahaba/Tushka and 16 Holocene Upper and Lower Nubian samples. Using the mean measure of divergence (MMD)—a distance statistic with a significance test—inter-sample affinities will help detect the most likely population hypothesis. That said, succession from A-Group, and probably earlier Neolithic, through to Christian populations is not disputed in any studies. The contention is if/when discontinuity occurred. Thus, focusing on the two Late Palaeolithic, plus three recently recorded Neolithic samples (ca 5600–4300 BCE), from Al Khiday, El Ghaba, and R12 (table 1), may refine this potential timing.

Within this analytical framework, continuity versus discontinuity is restated as statistical hypotheses: H_0: there is no versus H_A: there is a significant dental morphological difference between late Pleistocene and Nubian Holocene samples. Of course, it is the contribution of evidence from multiple fields that will convince researchers, but this approach should provide useful insight. These results will in turn have ramifications regarding craniodental changes in the agricultural transition. Again, do they result from selection that affected other populations worldwide, albeit to a lesser degree, or something else?

2. Material and methods

Eighteen samples associated with subsistence strategies ranging from hunting–gathering to intensive agriculture [18,31,37,38] are listed in table 1. Because Gebel Sahaba/Tushka (‘Gebel Sahaba’ from here on) is ‘biologically indistinguishable’ from the less dentally complete Wadi Halfa sample ([11, p. 316]; [39–41]); (electronic supplementary material, Note S1), it is analysed in the latter's place. Detailed backgrounds are available in earlier articles [33,42–44], except the Al Khiday Late Palaeolithic and Neolithic samples described here. Archaeological evidence is provided elsewhere [45–47], but in brief, the Al Khiday cemetery was used for greater than 10 000 years by Late Palaeolithic through to Meroitic populations. Radiocarbon dating of the mineralized Late Palaeolithic skeletons is problematic, like similar-aged Lower Nubian sites [40,48]. However, recent indirect dating at Al Khiday supported a 12th millennium BCE association [49]. The Neolithic skeletons were also dated indirectly (5000–4300 BCE). A key discriminating factor is body position, with Late Palaeolithic extended and prone (electronic supplementary material, figure S1), and Neolithic burials flexed, as is common across Nubia.

The aim is to use ASUDAS traits, many known to be highly heritable, h² = 0.60–0.93 [50,51], as proxies for ancient DNA (aDNA). Owing to degradation from environmental heat [52,53], aDNA has not been recovered in the above samples. Thus, affinities from ASUDAS data were calculated using the MMD [54,55] to estimate genetic relatedness. In support, a Mantel correlation of 0.84 resulted between distance matrices from ASUDAS traits (MMD) and more than 350 000 single-nucleotide polymorphisms (i.e. Hudson F_st) in 12 matched North and sub-Saharan African samples [56].

The 36 traits (electronic supplementary material, table S1) were used in earlier affinity studies [22,23,33,55–58]. Recording entails using standardized examples of trait expressions on a rank-scale to address interobserver error [36]. The latter is not an issue here because all data were recorded by J.D.I. Other reasons for choosing these traits include minimal sexual dimorphism to allow sample pooling, preservation despite crown wear, and their conservative evolution is ideal for diachronic biodistance analyses [19,36].

The rank-scale data were first dichotomized into standard states of present and absent [36], to simplify the presentation of trait frequencies and as required for the MMD [54,55,59,60]. As a benefit, dichotomization can increase h² [61]. The MMD calculates phenetic affinities, where lower values indicate greater similitude and vice versa. The formula corrects for low (≤0.05) or high (≥0.95) trait frequencies and small samples (n < 10) [54,59]. To determine if samples differ significantly, MMDs are compared with their standard deviations (Sx), e.g. if MMD > 2 × Sx, the null hypothesis S₁ = S₂ (S = sample) is rejected at the 0.025 level [54,55].

While the MMD is a robust statistic [55], trait editing is recommended to increase precision. Minimally contributory traits should be deleted [60]. Some are obvious, e.g. mandibular torus frequency is 0.00 across samples (electronic supplementary material, table S1). Others are identified quantitatively, here using correspondence analysis (CA) to discern which traits vary least across samples in combination biplots [62,63]. Additional CA information is available elsewhere [33,58] (electronic supplementary material, Note S2). Further, traits highly correlated with others can cause unwanted differential weighting of underlying dimensions [54]. Therefore, Kendall's tau-b was used to identify those most often highly inter-correlated (τ_b ≥ |0.5|).

Finally, beyond MMD distance matrices, inter-sample variation can be visualized with multidimensional scaling (MDS). SPSS 26.0 Procedure Alscal produced spatial representations of the samples.

3. Results

The number of individuals and percentages expressing traits by sample reveals inter-sample uniformities, except Gebel Sahaba (electronic supplementary material, table S1). The latter exhibits relatively high frequencies of mass-additive traits [33]. This is supported by an initial 36-trait MMD comparison. In the matrix (electronic supplementary material, table S2) and three-dimensional MDS plot (electronic supplementary material, figure S2), Gebel Sahaba differs significantly from all others (MMD = 0.09–0.26), including Late Palaeolithic and Neolithic Al Khiday (0.12 and 0.17), plus the other new Neolithic samples from Upper Nubia (0.16 and 0.17).

Following editing, 15 traits were deleted (electronic supplementary material, Note S2, table S3, figures S3 and S4), leaving 21 (electronic supplementary material, table S4). The new distance matrix (electronic supplementary material, table S5) and MDS (figure 2) place Gebel Sahaba (GSA) even farther from the rest (MMD = 0.14–0.44); affinities among the latter changed minimally, including the new Neolithic samples (AKN, GHB, R12). Al Khiday Late Palaeolithic (AKH) is close to Hierakonpolis C-Group (MMD = 0.04), Al Khiday Neolithic (0.01), Neolithic Ghaba (0.06), Kerma Ancien/Moyen (0.02), Kerma Classique (0.01) and Napatan Tombos (0.04); these distances do not differ significantly (electronic supplementary material, table S6). Of all samples, Late Palaeolithic Al Khiday is farthest from Final Neolithic Gebel Ramlah (MMD = 0.19), Meroitic (0.17) and Gebel Sahaba (0.17). An improvement in MMD performance is supported by the MDS. Beyond Gebel Sahaba being farther from the remaining, more homogeneous samples, the latter reveal a Lower/Upper Nubian division indicative of a north-to-south cline in ASUDAS frequencies (see Discussion below).

Figure 2.

Three-dimensional MDS of 21-trait MMD distances among Nubian samples. Three-letter abbreviations from table 1 and text. The MDS Kruskal's stress formula 1 value = 0.151 and r² = 0.886. (Online version in colour.)

4. Discussion

Assuming phenetic affinities reflect genetic relatedness, Gebel Sahaba appears too divergent to be ancestral to succeeding Nubians—differing significantly based on 36 and 21 traits. Such findings were reported previously [22–30,33,34]. These same studies indicate the Gebel Sahaba/Tushka/Wadi Halfa population was not indigenous to Nubia or the region, instead showing affinities to sub-Saharan Africans, notably West Africa. This too is not new, and two earlier studies reported cranial similarities with sub-Saharan samples: West African Ashanti [41], and Late Palaeolithic Ishango, Democratic Republic of the Congo ([40], also see [64]).

This is expediently illustrated in a two-dimensional plot of 36-trait distances (figure 3) among Gebel Sahaba, Al Khiday, pooled Lower (LNU) and Upper (UNU) Holocene Nubian samples (from table 1), and 12 early Holocene through historic samples from West, Central, and East sub-Saharan Africa ([23,24,55,57], electronic supplementary material, Note S3, table S7). Of interest, the Ashanti crania from [41] comprise the Ghana (GHA) sample near Gebel Sahaba. The latter's location shows it most akin to West Africans and three Central African samples, sharing traits common among subcontinental populations [57,65,66]. None of these distances differ significantly (electronic supplementary material, table S7). Except Gebel Sahaba, inter-sample distances parallel geographical locations (also [23,55,56]), where dimension 1 approximates west-to-east and dimension 2, north-south.

Figure 3.

Two-dimensional MDS of 36-trait MMD distances among Late Palaeolithic Gebel Sahaba (GSA) and Al Khiday (AKH), pooled Lower (LNU) and Upper (UNU) Holocene Nubian samples from the present study, and 12 early Holocene and historic samples from West, Central and East sub-Saharan Africa (details in the electronic supplementary material, Note S3 and table S7). MDS Kruskal's stress formula 1 = 0.214 and r² = 0.787. (Online version in colour.)

A sub-Saharan population in late Pleistocene Nubia should not be unexpected, given northward expansions of Sahelian vegetation and sub-Saharan fauna during Saharan ‘green’ periods; the most recent initiated 15 000 BP [67], before its maximum around 9000 BP [67–69]. It may seem surprising that these apparent migrants originated so far away, but many well-watered migration routes were available then [22,26,68]. In any event, information on biological distinctiveness and non-local derivation is not novel, as mentioned. Nevertheless, diachronic change in a continuous, geographically stable Lower Nubian population from the late Pleistocene onwards is still proposed as a viable explanation [3].

What is new, however, is the 12th millennium Al Khiday sample. None of the crania has been reconstructed but they appear robust (electronic supplementary material, figures S5 and S6), perhaps not unlike contemporaneous Lower Nubians, northwest African Iberomaurusians [29] or Central African Ishango [64]. Odontometrics have also not been recorded, but all teeth appear much larger than more recent samples, again not unlike the above material [27]. Yet, compared to Gebel Sahaba, Al Khiday teeth are simpler like in Holocene Nubians (electronic supplementary material, table S5; figure 2). In particular, distances with the Hierakonpolis C-Group and five Upper Nubian samples do not differ significantly. However, Al Khiday also expresses traits indicative of sub-Saharan origin ([57,65,66]; electronic supplementary material, Note S3), but like geographically proximate East Africans and one Central African sample (electronic supplementary material, Note S3, table S7; figure 3). Site location is a likely factor, as during humid periods it would literally have been in ‘sub’-Saharan Africa. Of course, regardless of climate, the Nile and tributaries acted as north-south migration routes during the whole of prehistory. Evidence of exchange is seen by an increase in sub-Saharan-like traits between Lower and Upper Nubian Holocene samples (figure 2; electronic supplementary material, table S1, Note S3).

On the above bases, selection did not account for craniodental changes between the Lower Nubian late Pleistocene and Holocene samples studied by continuity proponents. The Wadi Halfa/Gebel Sahaba/Tushka population, whether in residence for a few generations or a thousand years, contributed little, if anything, to the Holocene gene pool for in situ evolution to occur. However, a candidate ancestral population was present. While divergent from some, Late Palaeolithic Al Khiday is closer to all samples than Gebel Sahaba (electronic supplementary material, table S5). From this population then, craniofacial reduction relative to the masticatory-functional hypothesis [8,9] cannot be ruled out, given indications of Al Khiday robusticity. Neither can selection for size reduction in all teeth following the Pleistocene [1,16,18]. However, the lack of reduction in dental morphological complexity does not support in situ caries selection [7,20] in this Upper Nubian scenario.

In summary, the most parsimonious explanation is ancestors of Holocene agriculturalists were in Nubia—just not at Wadi Halfa, Gebel Sahaba, and Tushka. Although cultural diffusion with the incorporation of non-local resources occurred [70,71], with perhaps some immigration, it is unnecessary to hypothesize a significant post-Pleistocene influx of agriculturalists. The results suggest most future Nubian agriculturalists were in residence the entire time, though previously in the guise of Neolithic agro-pastoralists and intensive collectors. It would seem likely that, soil deflation aside, more Late Palaeolithic skeletal remains akin to Al Khiday may yet be discovered, possibly including Lower Nubia. So, long-term population continuity appears likely after all, perhaps including in situ selection for a reduction in cranial robusticity, as well as dental size (only), during the transition from hunting–gathering to agriculture.

Data accessibility

Nubian nonmetric data are listed in the electronic supplementary material, table S1. Nonmetric data for the 12 sub-Saharan African samples in Irish [23,24,55,57].

Authors' contributions

J.D.I.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, visualization, writing—original draft, writing—review and editing; D.U.: funding acquisition, investigation, validation, writing-review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Competing interests

We declare we have no competing interests.

Funding

Funding for J.D.I. from the National Science Foundation (grant nos. BNS-9013942 and BNS-0104731), American Museum of Natural History, Institute for Bioarchaeology, and Institute of Archaeology and Ethnology, Polish Academy of Sciences. Funding for D.U. Sudanese fieldwork from MAECI and Universities of Padova, Milano and Parma.