While the term ‘Neolithic’ remains ambiguous and this period occurred at different times worldwide, it generally implies the domestication of wild animals and plants, as well as the adoption of sedentary settlements1,2. The transition from hunting-gathering economies to agriculture-based ones, also known as Neolithization, is one of the most important dietary revolutions in human history3,4. Beyond being a revolution, a progressive intensification of plant consumption is believed to have begun long before domestication in the Neolithic1,5. Evidence of an early shift to grain-based resources is demonstrated by the discovery of a substantial archaeobotanical assemblage in the Upper Palaeolithic site of Ohalo II, in the Near East, dated to approximately 23,000 cal BP5 (Fig. 1). This transformation intensified with the Natufians, a hunter-gatherer group that inhabited the Near East during the Late Pleistocene and the beginning of the Holocene (14,600–11,500 cal BP)6. A shift towards an increased reliance on plant foods occurred during this period6,7,8, probably driven by several factors, including the depletion of large game species and the availability of a wider range of edible plants in the environment, which led to the adoption of a broad-spectrum diet9. Natufian hunter-gatherers also engaged in early forms of plant cultivation, such as the intentional planting and harvesting of wild cereals. This practice probably paved the way for the development of agriculture in the region10,11.

Fig. 1: Location of the Taforalt site in Morocco and the other sites mentioned in the text.
figure 1

The circles indicate Iberomaurusian sites, the squares indicate European Upper Palaeolithic sites, the triangle indicates the Natufian site and the star indicates the Neolithic site in the Levant.

The preconditions of the transition to food production in the Levant are deeply rooted in the Natufian hunter-gatherers, but this transition is still a poorly understood and complex phenomenon in northwest Africa12. In this region, a shift towards a reliance on plant resources in the diet was thought to be a relatively late phenomenon, which started with the spread of domesticated species from the Near East into this region during the Neolithic (~7,600 BP)12,13,14. In recent years, scholars have become increasingly interested in whether the Iberomaurusians, a population with some genetic connections with the Natufians15, exhibited changes that preceded the transition to farming in North Africa16,17. Recent investigations at the site of Taforalt (Fig. 1), Morocco have suggested early consumption of carbohydrate-rich plants associated with the Iberomaurusian culture. This has been attested by the high number of wild plant taxa along with the prevalence of tooth caries among the human burials17.

The Iberomaurusian hunter-gatherers, characterized by bladelet-based technology, inhabited North Africa during the Late Pleistocene. The first evidence of this culture, found in Tamar Hat (Fig. 1), dates back to 25,000 cal BP18,19. The timing of its end remains uncertain, with some evidence suggesting the possibility of its persistence into the Holocene after 11,000 cal BP20,21.

Two key areas of interest are the domestication of plant and animal species, a crucial step in agricultural development, and the adoption of a sedentary lifestyle, often associated with plant cultivation. While there is no evidence of local domestication during the Iberomaurusian period22,23,24, some behaviours suggestive of a shift towards sedentism in the subsistence economy were present among these hunter-gatherers. For example, at the Iberomaurusian site of Taforalt (Fig. 1), evidence points to the selective harvesting and possible storage of some edible plant species17. This is documented by the presence of fragments of alfa grass (Stipa tenacissima), which would have been used to make baskets. Wild plants have been recovered from other Iberomaurusian sites and could have been collected for the purpose of consumption, such as at Tamar Hat, Algeria25, and Ifri el Baroud, Morocco (Fig. 1)26.

Currently, our knowledge of the Iberomaurusian diets is mostly derived from zooarchaeological evidence. Studies have revealed that the Iberomaurusians relied primarily on ungulates, mainly represented by the Barbary sheep (Ammotragus lervia), in addition to snails24,27. These conclusions find further support in an isotopic study conducted on bulk collagen, which identified a predominance of meat in the diet of the Taforalt humans28. Studies on the exploitation of marine resources for food are scarce despite both the proximity of Iberomaurusian sites to the coast29 and the recovery of marine mollusc shells from various Iberomaurusian sites, where these shells appear to have been used for ornamental purposes29.

However, it is worth noting that the faunal remains may not fully represent the entire spectrum of the foods consumed. This limitation arises because plant remains are less likely to preserve well in the archaeological record, and their recovery and identification may not be as frequent as that of animal bones30,31,32. Furthermore, the detection of plant consumption can be easily overprinted by the presence of meat consumption when assessed using nitrogen isotopes on bulk collagen33. In terms of settlement patterns, while no stone-built structures similar to those in Natufian settlements are evident5, the presence of large Iberomaurusian cemeteries (such as Taforalt and Afalou; Fig. 1) in frequently reused sheltered sites—from 15,000 to 13,000 cal BP (Fig. 1)17,34—is interpreted as evidence of sedentarism35.

Taforalt is one of the two largest known Iberomaurusian cemeteries. This site has yielded substantial amounts of recovered plant remains. In addition, it contains the longest and best-dated occupation sequence for the Iberomaurusian period35,36,37. To date, it is one of the oldest cemeteries in North Africa, with the largest number of human burials (including adults, adolescents and infants). The human remains were directly dated to 15,077 to 13,892 cal BP17, which coincides with a rapid warming period following the Last Glacial Maximum26. It is a key site for studying human dietary behaviour during the Late Pleistocene in North Africa and offers an exceptional opportunity to investigate human dietary behaviours at the end of the Late Pleistocene and before the spread of farming practices in the region. In addition, we have at this site contradictory evidence of dietary reliance on meat (faunal remains24, C and N isotopes28) and plant foods (plant remains, tooth caries17). A plant-based diet combined with economic intensification could indicate a transitional subsistence strategy towards sedentism. By combining previously used isotope tracers and new ones that are more sensitive to plant consumption, we aimed here to investigate the dietary habits and the mobility patterns of pre-Neolithic hunter-gatherers in North Africa at Taforalt. In particular, we investigated the proportion of plants in their diet and whether this population was relying on local foods.

To accomplish this, we evaluated the bulk stable isotope compositions of carbon (δ13Ccollagen) and nitrogen (δ15Ncollagen) in bone and dentine collagen to reconstruct dietary patterns of both human individuals and coexisting fauna rather than to determine the presence/absence of food products in the diet of a population (Supplementary Information Section 2 and Supplementary Fig. 1). However, these bulk isotopic results can be impacted by baseline variations related to environmental parameters such as aridity, essential element availability or the nature of local mycorrhizae38,39,40. To overcome this issue, compound-specific isotope analysis of single amino acids (CSIA-AA) is used to determine more precisely the trophic position (TP) of an organism independent of environmental factors using the δ15N results for two amino acids: Phe and Glu38. In addition, δ13C analysis of amino acids such as Phe and Val can effectively distinguish between four main dietary groups (C3, C4, marine and freshwater) (Supplementary Information Section 2)41.

While organic isotopic proxies are powerful for dietary reconstruction, their application in Africa often faces challenges due to limited collagen preservation in fossil remains from arid environments42. Hence, we enhanced our analysis by investigating zinc isotope ratios (δ66Zn) in tooth enamel, a method that has been proved to reliably document trophic levels43,44,45 even in the absence of collagen preservation43,46,47. Zinc and nitrogen isotope ratios have an inverse relationship, wherein lower δ66Zn values reflect an elevation in the TP. Given that baseline effects related to geological and environmental parameters can influence δ66Zn values43,46,48, we ensured valid comparisons by also conducting analyses of commonly used mobility indicators, including strontium (87Sr/86Sr) and sulfur (δ34S) isotope ratios49,50. All details on these isotopic proxies are provided in Supplementary Information Section 2.

We analysed the human remains from the Iberomaurusian burials recovered from sector 10 and associated fauna at Taforalt (Supplementary Information Section 1 and Supplementary Tables 13). Human samples consisted of 25 teeth (permanent and deciduous) and seven bone samples belonging to seven identified and ten unassigned individuals (Supplementary Tables 3, 14 and 23). The tissues sampled record different periods of the lives of the individuals, including the breastfeeding period (Supplementary Table 15). Special attention was therefore paid to the potential impact of breastmilk consumption on the isotope ratios throughout the text and figures (Figs. 2 and 3 and Supplementary Information Section 5). To preserve morphometric information, the human teeth samples were CT-scanned, and we took this opportunity to document the presence or absence of hypoplasia and caries (Supplementary Information Sections 3 and 7 and Supplementary Table 4). We selected several teeth and bones from various species of associated faunal taxa (sectors 8 and 10) that were exploited by humans to reconstruct the isotopic baseline for Taforalt24 (nsamples = 20; Supplementary Fig. 1 and Supplementary Table 13): Barbary sheep (Ammotragus lervia), Equidae (Equus sp.), hare (Lepus sp.), hartebeest (Alcelaphus buselaphus), gazelle (Gazella sp.) and Rhinocerotidae. We also analysed two canid specimens (Canis sp. and Vulpes vulpes) to evaluate isotope values associated with a meat-based diet51. The faunal taxa were identified using traditional zooarchaeological methods24 and zooarchaeology by mass spectrometry (ZooMS)52,53 (Supplementary Information Section 3 and Supplementary Table 2). Through the use of these isotopic proxies, the focus of this work is to quantify this population’s reliance on plants and determine whether their transition to a more plant-based diet mirrors that of the Levantine Natufian.

Fig. 2: Isotopic ratios of various elements from the human and faunal teeth/bone of Taforalt.
figure 2

a, Carbon (δ13C) and nitrogen (δ15N) isotopic ratios from bulk collagen of dentine and bone samples. Each point corresponds to a sample; samples from the same individual are connected with a line. The typical analytical error is 0.1‰ for the two isotope systems. VPDB, Vienna PeeDee Belemnite; AIR, atmospheric N2; Ind, individual. b, Zinc (δ66Zn) and strontium (87Sr/86Sr) isotopic ratios from enamel bioapatite. Each point corresponds to a sample; samples from the same individual are connected with a line. The typical analytical error is 0.05‰ for δ66Zn and 7 × 10−6 for 87Sr/86Sr. c, Carbon (δ13C) and nitrogen (δ15N) isotopic ratios from bulk collagen of dentine and bone samples with associated 95% confidence ellipses. Each point corresponds to the average value of all samples coming from a single individual (nindividual = 33; 44 samples in total); the error bars give the standard deviation for all the values from the same individual. d, Zinc (δ66Zn) and strontium (87Sr/86Sr) isotope ratios from enamel bioapatite with associated 95% confidence ellipses. Each point corresponds to the average value of all samples coming from a single individual (nindividual = 33; 41 samples in total); the error bars give the standard deviation for all the values from the same individual.

Fig. 3: Zinc and nitrogen isotope values versus TP.
figure 3

a, Zinc (δ66Zn) isotope values versus the TP obtained from single amino acids (Supplementary Information Section 4). b, Nitrogen (δ15N) isotope values from bulk collagen versus the TP obtained from single amino acids. The TP was estimated from δ15NPhe and δ15NGlu values (Supplementary Information Section 2). Samples from the same human individual are connected with a line.

Results and discussion

The measured δ66Znenamel, δ13C, δ15N (bulk collagen and amino acids), 87Sr/86Srenamel and δ34Scollagen for the humans and fauna from Taforalt are presented in Figs. 24, Extended Data Figs. 13, Supplementary Information Section 4 (Supplementary Figs. 312 and Supplementary Tables 512), Supplementary Tables 1624 and Supplementary Fig. 16. The diets of the humans are discussed in Supplementary Information Section 5 (Supplementary Figs. 1315 and Supplementary Table 13).

Fig. 4: Measured δ15NPhe and δ15NGlu values on human and faunal collagen from Taforalt.
figure 4

a, Values according to the formation time of the sample. b, Values according to the species. The dashed black lines indicate approximately the theoretical TP of herbivores (TP = 2) and carnivores (TP = 3). The dashed grey line is the intermediate (TP = 2.5). Samples from the same human individual are connected with a line. RB, rib bone; LB, long bone; dm2, deciduous second molar.

All the faunal remains from Taforalt exhibit similar 87Sr/86Srenamel values to the humans, which are close to the modern seawater value (~0.7092, Fig. 2b)49. Since the herbivores also exhibit this seawater value and given the δ13Ccollagen of human individuals, it is unlikely that the similar values in the humans indicate marine food consumption; rather, they probably reflect the values of the local geology, which is dominated by calcareous bedrocks35 (expected to be between 0.707 and 0.709 (refs. 49,54); Fig. 2b). All of the other proxies used in this study (δ15Ncollagen, δ13Ccollagen, δ34Scollagen, δ15NAA and δ13CAA) suggest the absence of regular aquatic food consumption (Supplementary Information Section 4).

Trophic level information was determined using three isotopic tracers: δ66Znenamel, δ15Ncollagen and the TP (C3) equation based on δ15NPhe and δ15NGlu values38,55 (Fig. 3 and Supplementary Information Section 4). Given that the canids from Taforalt primarily have a meat-based diet51 and that the plant portion of their diet consists of fruits, a resource showing exceptionally low zinc concentrations56, their δ66Znenamel values should be indicative of a carnivorous diet. For δ66Znenamel, the trophic level spacing (TLSherbivores–canids) for all individuals is +0.62‰ and +0.70‰ if we consider only teeth formed post-weaning. This is close to that of Late Pleistocene sites in Laos (+0.63‰)46,48 and higher than in a modern food web in Kenya (+0.40‰)57. However, the TLSherbivores–canids should be considered with precaution, given the small number of canid samples and the fact that the teeth might have been impacted by consumption of their mother’s milk. When only considering human teeth formed after weaning, we found elevated adult human δ66Znenamel values, which indicates a low trophic level (0.78 ± 0.07‰, nsamples = 28, nindividuals = 12), and these values are close to those of Taforalt herbivores (nherbivores = 7, nsamples = 20, TLShumans–herbivores = +0.34‰). The offset is 0.32‰ between the humans and the Barbary sheep, the primary source of game at Taforalt24. In contrast, this isotopic spacing is much higher at other Late Pleistocene sites such as Tam Pà Ling (TLShumans–herbivores = +0.48‰) in Laos46,48, Gabasa (TLShumans–herbivores = +0.85‰) in Spain43 and the medieval site of Rennes (TLShumans–herbivores = +0.63‰) in France44. Furthermore, the δ66Znenamel results from the Taforalt humans overlap with those from populations with historically documented cereal-based diets and not with those from populations that regularly consumed meat (Supplementary Fig. 17)44, although this comparison does not consider baseline effects (Supplementary Information Section 4). As dietary zinc is likely to be primarily absorbed from animal sources58,59, the minimal isotopic differences between Taforalt Iberomaurusians and herbivores at low trophic levels (TLShumans–herbivores) and the elevated δ66Znenamel values provide compelling evidence of substantial plant consumption. This, in turn, affirms their meat intake as well.

This interpretation of δ66Znenamel data is also supported by the trophic level estimations obtained from the isotopic analyses of amino acids. The TP of adult humans, in tissues formed post-weaning, was found to vary between 2.2 and 2.6 with an average of 2.4 ± 0.2 (nsamples = 9). Thus, for the majority of individuals, plant resources were the primary source of dietary proteins. This finding highlights a substantial consumption of plant protein in a pre-agriculturist human population60,61. In particular, these TP values at Taforalt are similar to the TP values of Neolithic farmers from the Levant (Tell El Kerkh) (Fig. 1)62. Evidence for substantial plant consumption has also been found for two early modern humans (TP values of 2.5 and 2.6) at the Palaeolithic site of Buran Kaya in Crimea (Fig. 1), and this was similar to most of the associated canids at this site60. While the canids at Taforalt may not be categorized as pure carnivores, their TP values remain notably high, especially in the case of the red fox (Vulpes vulpes) with a TP of 2.9. It is interesting to note that the canids’ TP values surpass that of the humans, further supporting the idea that humans had a low reliance on animal protein. In particular, the TP results for individual 1 and an unassigned tooth closely resemble those of the herbivores (Fig. 4).

In addition to the small TLShumans–herbivores values for δ66Zn (0.34‰) and the low TP calculated by CSIA-AA, the δ15Ncollagen values between the humans and herbivores (Δ15N = 2.5‰) are smaller than those from other Upper Palaeolithic sites in Europe and Asia (Fig. 1) where animal proteins were the main dietary component (for example, Buran Kaya (Δ15N = 6.2‰), Oase (Δ15N = 10.8‰), Brno-Francouzska (Δ15 = 7.1‰) and Tianyuan (Δ15N = 6.4‰))60,63,64. Our results on the TLS between humans and herbivores are different from the Δ15N (TLS) observed by Lee-Thorp et al. (+4.2‰) for Taforalt (Supplementary Information Section 4 and Supplementary Table 10)28. While their study focused on the Barbary sheep, the primary hunted faunal species at the site24, it is important to consider that this species has a flexible diet65, which might have influenced the accuracy of the TLS value due to potential differences in isotopic baselines. Our study demonstrates that this species had variable δ15Ncollagen values while having a stable trophic level of 2.1 ± 0.0 based on δ15NAA values (Supplementary Information Section 4). However, Lee-Thorp et al.28 observed an absence of aquatic food consumption, aligning with our conclusions. In addition, Hedges and Reynard33 found that δ15Ncollagen bulk-based diet reconstructions tend to overestimate animal protein intake by 60–80% when a nitrogen isotopic ratio enrichment of 3‰ or more is applied using the standard model for δ15N interpretation. This conclusion is supported by the association of a TLS of +3‰ with a plant intake of 50% (ref. 66) and a TLS of +4‰ among European Neolithic farmers with a meat intake of 40% (ref. 67). We should take into consideration that the plants eaten by humans could be more enriched in δ15N than plants consumed by herbivores due to the effect of charring, which can increase their δ15Ncollagen by up to +1‰ (ref. 68). The δ15Ncollagen values observed in humans are probably affected by their consumption of these processed plants69,70, compared with unprocessed forage plants consumed by herbivores. Our TLS estimations for Taforalt based on δ15Nbulk of +4.2‰ and +2.5‰ could therefore suggest a plant food intake of about 50% in the Taforalt human diets. This is in agreement with our conclusions based on Zn isotope ratios and CSIA-AA, the presence of a variety of wild plants at the site17 and the high prevalence of tooth caries and other periodontal diseases, which frequently exceeds those observed for hunter-gatherers, all suggesting a high consumption of fermentable starchy plants4,17,71. However, it must be stressed that the Taforalt humans studied here were not strict vegetalians, as isotopic offsets between the δ15N and δ66Zn herbivore and human values are documented and because zooarchaeological data indicate that animal protein was consumed. In particular, cut marks were observed on the faunal assemblage, mostly on Barbary sheep but also on gazelle, equid, large bovines and hartebeest24. These cut marks provide evidence of butchery and processing of animal remains, which directly supports the notion that animal protein was an integral part of the Taforalt human diet.

On the basis of multiple isotope proxies, we can also document an early weaning age for an infant (Ind-6) at Taforalt (Fig. 4 and Supplementary Information Section 5). The noticeable decline in TP, as calculated from the δ15N values of single amino acids, among tissues with varying formation periods is particularly evident. The tissues, such as the deciduous second molar (TP = 2.8, formed over a span of approximately −0.34 to 1 year72,73) and the long bone (TP = 2.9, formed over the first year72,73), which have slower remodelling rates, exhibit higher TP values. In contrast, the rib bone, with its faster turnover rate68 and TP of 2.5, is likely to have recorded dietary information much closer to the time of the individual’s death (6–12 months73) (Supplementary Information Sections 24). This pattern of decreasing TP values strongly suggests a rapid transition in the individual’s diet, with the introduction of adult foods playing a substantial role in this dietary shift74 (Fig. 4).

This is evidence that weaning was initiated before 1 year of age and possibly with plant-based foods, since we observed a clear decrease in this individual’s TP (2.8 to 2.5). Unlike at other sites75, we do not see a clear weaning pattern in the δ66Zn results when comparing different teeth of a single individual or at the population level (Supplementary information Section 5 and Supplementary Fig. 13). This observation may be due to a sample bias, as the limited sample size per individual prevented the tracing of potential weaning patterns. Alternatively, this could be attributed to the early introduction of solid foods in infant diets. The adoption of a starchy diet in Taforalt may have facilitated early weaning, a pattern commonly associated with the transition to agriculture due to the availability of soft and digestible foods such as cereals. However, early weaning can result in increased stress and mortality for infants76. This contrasts with hunter-gatherer societies, where extended breastfeeding periods are the norm due to the limited availability of suitable weaning foods4,77. These observations suggest that changes in diet and lifestyle in the Iberomaurusians from Taforalt might have had important impacts on infant feeding practices. However, it is clear that additional detailed analyses are needed to fully understand this weaning pattern on a larger scale.

According to the broad-spectrum and dietary breadth models, a reduction in the availability of large to medium-sized game animals often leads to increased foraging for previously overlooked resources such as lagomorphs and small birds and an increased exploitation of wild plants10,78. This hypothesis has been commonly applied to explain the emergence of farming in Southwest Asia, where the Natufian hunter-gatherers, initially reliant on small to medium-sized ungulates, adapted their subsistence strategy due to ecological pressure on these animals79. As a result, they gradually diversified their diet by incorporating a broader range of food resources, including wild plants. This may have been the case for the Taforalt population, as evidenced by the high incidence and diversity of charred macrobotanical plant remains found in the Grey Series level17. The prevalence of caries in the human teeth in burials also suggests a substantial reliance on highly cariogenic wild plant foods such as sweet acorns, pine nuts and some legumes. Furthermore, the presence of grinding stones in the same deposits suggests plant processing, which is possible evidence that the nuts and acorns were ground into flour or meal17. While the removal of the central upper incisors was a prevalent practice among 90% of the Taforalt population and is common among the Iberomaurusians17,79, it is important to note that this practice is not linked to oral pathology. Instead, it may have impacted the functional use of teeth for mastication79.

The δ13C amino acid results presented here indicate that most of the humans and herbivores consumed C3 plants (Supplementary Information Section 4), which is the photosynthetic pathway of all the edible plant species found at Taforalt. It is likely that most of the wild plants were collected during autumn, such as acorns, while pulses were harvested from late spring to summer17. The inhabitants probably stored plants, which would have ensured consistent food staples throughout the year17.

These findings suggest a notable increase in the reliance on plant resources by the Taforalt population. While there is no evidence of a decline in Barbary sheep (the main hunted species during the Iberomaurusian period24) at the site, it is plausible that the seasonal availability of these species and other ungulates at the site influenced the access to meat proteins through the year. The mortality age of Barbary sheep and gazelle points to hunting activities occurring between spring and early summer24. Simultaneously, the increased abundance of wild plant resources in the inhabitants’ environment may have played a role in their subsistence strategy behaviour. Land snails might have been consumed seasonally too27. The consumption of wild plant resources (such as acorns) may explain why most of the Iberomaurusian sites were located in the coastal Mediterranean forest regions of Northwest Africa. However, more Iberomaurusian sites need to be studied to confirm this hypothesis.


Our study highlights the importance of the Taforalt population’s dietary reliance on plants, while animal resources were consumed in a lower proportion than at other Upper Palaeolithic sites with available isotopic data. The potential early weaning of infants at Taforalt reinforces the notion of a plant-based food focus for the population, potentially extending to the primary source of nutrition for infants. However, it is crucial to acknowledge that further comprehensive investigations are necessary to fully understand these findings and their implications. Evidence of intensive exploitation of wild plants at the end of the Late Pleistocene is also documented in the Near East with the Natufian hunter-gatherers, who developed cultivation and became some of the earliest agriculturists. In that region, it is believed that the Younger Dryas climatic deterioration in the early Holocene (11,000–10,300 uncal BP) was the major trigger for systematic cultivation in response to the reduction of the vegetal cover and, consequently, the availability of exploited wild plants8,80. Although the Natufian and Iberomaurusian populations had broad similarities regarding the preconditions for the emergence of food production (intensive plant consumption and increased sedentism) and genetic connections (63% of shared genes between Natufian and Iberomaurusian individuals)15, these factors did not lead in North Africa to a similar local development of agriculture and farming despite the high reliance on plants as a food staple during the Later Stone Age. While the origin of this dissimilarity is still open to debate, the Younger Dryas cooling phase might have reduced the abundance of plant resources, which could explain why Iberomaurusian sites became less occupied during the period81.


Zinc and strontium isotope analyses

Zinc and strontium were extracted from tooth enamel. The sampling strategy details are given in Supplementary Information Section 7 and Supplementary Tables 16 and 17. To explore potential dietary variations during tooth development in humans, we used a specific sampling approach involving the collection of enamel samples from both the upper (top) and lower (bottom) portions of the tooth crown (Supplementary Table 16 and Supplementary Information Section 7). The extended formation period of enamel means that it captures dietary information from various stages of an individual’s life. By sampling both the top and bottom parts of the crown, we aimed to examine potential dietary changes and isotopic variations that could provide insights into the individual’s nutritional history over time. It is important to acknowledge that this approach was not uniformly applied to all teeth due to the varying condition of the teeth themselves. Some teeth presented a challenge due to insufficient enamel (the material we also used for Sr analyses) extending along the height of the crown. This deficiency in preserved enamel was primarily attributed to wear and abrasion, affecting teeth such as SEVA 35976, 35968, 35959 and 35973. Enamel samples were collected using a precise and minimally destructive method. We used a diamond-tipped burr to mechanically extract a powder from multiple parts along the height of the tooth crown. The amount of enamel collected ranged from 5 to 20 mg, ensuring that we captured sufficient material for isotopic analyses while minimizing the destructive sampling as much as possible.

For faunal teeth, particularly those of herbivores, we implemented a multisampling method to account for potential isotopic variations (Supplementary Fig. 2 and Supplementary Table 17). Faunal taxa such as equids, Barbary sheep, hartebeest and rhinoceros with teeth of sufficient height provided an opportunity for multisampling. We collected samples from the tooth crown at multiple points, approximately three to five times along its height (5 to 20 mg), to track isotopic changes from the estimated initiation of enamel formation to the completion of the crown (Supplementary Table 17). To account for the influence of nursing signals on isotopic signatures, we also estimated the age of enamel formation for the faunal species (more details are provided in Supplementary Information Section 3).

The Zn and Sr samples were prepared from separate enamel powder aliquots. For each column chromatography batch, we included a matrix-matched powder standard with known isotopic composition (SRM 1486 for 87Sr/86Sr and SRM 1400 for δ66Zn) and a procedural blank to monitor contamination. The purification step of zinc was achieved using the ion exchange method. The samples were dissolved in 1 ml of HCl, evaporated and then dissolved in 1.5 M HBr. Zinc was purified following the protocol adapted from Moynier et al.82 and described in Jaouen et al.57. The purification of Zn was achieved on HNO3-cleaned AG1X8 resin conditioned with 1.5 M HBr, which was also used for matrix elution. For the Zn elution, we used 5 ml of 3% HNO3. The Zn isotope ratios were measured using a Thermo Neptune Multicollector inductively coupled plasma mass spectrometer at Max Planck Institute for Evolutionary Anthropology (Leipzig, Germany). Each enamel sample was analysed in duplicate. The average values were calculated from the duplicate measurements with a standard deviation of 0.0‰ to 0.05‰. Standard sample bracketing and Cu doping procedures were implemented to enhance measurement accuracy83. All the zinc data are reported using the δ notation in per mille relative to the standard JMC-Lyon. The typical analytical standard deviation of the bracketing standard for δ66Zn is 0.00‰ to 0.03‰. The sample Zn concentrations were estimated by measuring the 64Zn signal intensity of three solutions with known concentrations (150 ppb, 300 ppb and 600 ppb) relative to the sample concentrations.

For strontium, the purification was achieved using the protocol adapted from Copeland et al.84. The enamel was dissolved in 2 ml of HNO3, evaporated and subsequently re-dissolved in 3 M HNO3. Strontium was purified using Sr.spec.TM resin and eluted from the resin with 1.5 ml of ultrapure deionized Milli Q water before evaporation. The dry samples were dissolved in 2 ml of 3% HNO3 for the isotopic measurements. The Sr isotopic ratios were measured on the Thermo Neptune Multicollector inductively coupled plasma mass spectrometer. The interference from Rb with 87Sr/86Sr measurements was corrected on the basis of the repeated analysis of the standard (SRM 987). The Sr concentrations were calculated by measuring the 88Sr signal intensity (V) of three diluted solutions with known concentrations (100 ppb, 400 ppb and 700 ppb).

Carbon and nitrogen isotope analyses

For collagen extraction, both human and faunal tooth dentine and bone samples, weighing between ~200 and 500 mg, were carefully removed using a diamond-tipped burr. Collagen was extracted following the protocol described in Talamo and Richards85. Dentine and bone were cleaned by abrasion and then demineralized in HCl for several weeks at 4 °C. The samples were immersed in NaOH for 30 minutes. The purified solution containing collagen was rinsed and soaked again in HCl. The insoluble collagen was solubilized in HCl (pH 3) at 75 °C for 20 h. The supernatant containing the collagen was filtered with Ezee filters and then ultra-filtrated to collect the >30 kDa collagen molecules, frozen for 48 h and lyophilized.

The C and N isotopic ratios were obtained using a Thermo Finnigan Flash EA coupled to a Delta V isotope ratio mass spectrometer. Carbon and nitrogen stable isotope ratios were measured relative to Vienna PeeDee Belemnite and atmospheric N2, respectively. The analytical errors of 0.1‰ and 0.2‰ for δ13C and δ15N were determined by the repeated analysis of internal and international standards (IAEA-N1, IAEA-N2, MET and MRG).

Sulfur isotope analyses

The sulfur isotopic analyses were conducted by Isoanalytical. Eight milligrams of collagen were loaded into tin capsules in addition to the NBS-1577B (bovine liver powder) standard used as a control. The samples and the standards were measured using an EA-IRMS (ANCA-GSL/20-20, Europa Scientific), and the isotope ratios are reported to the international reference standard Vienna-Canyon Diablo Troilite.


CSIA-AA was conducted in the commercial Stable Isotope Facility of the University of Davis, California. The samples were analysed following the protocol developed by Yarnes and Herszage86. The detailed procedure is described at


Faunal remains were taxonomically identified using both traditional zooarchaeology24 and ZooMS52 (Supplementary Tables 2 and 17). ZooMS is a minimally destructive proteomic method that focuses on unidentifiable bone fragments. Through the analysis of bone collagen protein type I, it provides a taxonomic identification based on protein amino acid sequence variation. ZooMS analysis followed extraction protocols detailed elsewhere52,53,87. Selected faunal specimens from Taforalt were analysed through ZooMS and sampled using pliers (10–20 mg). Soluble collagen was first extracted through incubation in 100 µl of 50 mM ammonium bicarbonate (NH4HCO3, pH 8.0) buffer at 65 °C for 1 h. To improve the taxonomic identity obtained from soluble collagen, each bone sample was demineralized in 250 µl of 0.6 M hydrochloric acid (HCl) at 4 °C for 20 h. The samples were then centrifuged for 1 min at 9,520 g, and the supernatant was removed. The demineralized collagen was rinsed three times in 200 µl of 50 mM ammonium bicarbonate to be neutralized to pH 8, and 100 µl of 50 mM ammonium bicarbonate was added to each sample. Next, the samples were incubated at 65 °C for 1 h. Then, 50 µl of the resulting supernatant was digested with trypsin (0.5 µg µl−1, Promega) at 37 °C overnight, acidified using 1 µl of trifluoroacetic acid (20% TFA) and cleaned on C18 ZipTips (Thermo Scientific). Digested peptides were spotted in triplicate on a MALDI Bruker plate with the addition of α-cyano-4-hydroxycinnamic acid (Sigma) matrix. MALDI-TOF-MS analysis was conducted at the Fraunhofer IZI in Leipzig, Germany, using an autoflex speed LRF MALDI-TOF (Bruker) in reflector mode, with positive polarity, with matrix suppression up to 590 Da and collected in the mass-to-charge range 700–3,500 m/z. Triplicates were merged for each sample, and taxonomic identifications were made manually through peptide marker mass identification in comparison to a database of peptide marker series for medium- to larger-sized Pleistocene mammalian species53,88. We performed laboratory blanks alongside the samples to assess any potential contamination by non-endogenous peptides. These remained empty of collagenous peptides, excluding the possibility of modern laboratory or storage contamination.

Statistical analyses

Statistical analyses and data visualization were performed using R software (R version 4.3.1). The function stat_ellipse in the R package ggplot2 was used to produce ellipses with 95% confidence (

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.