The Earliest Horse Harnessing and Milking

See allHide authors and affiliations

Science  06 Mar 2009:
Vol. 323, Issue 5919, pp. 1332-1335
DOI: 10.1126/science.1168594


Horse domestication revolutionized transport, communications, and warfare in prehistory, yet the identification of early domestication processes has been problematic. Here, we present three independent lines of evidence demonstrating domestication in the Eneolithic Botai Culture of Kazakhstan, dating to about 3500 B.C.E. Metrical analysis of horse metacarpals shows that Botai horses resemble Bronze Age domestic horses rather than Paleolithic wild horses from the same region. Pathological characteristics indicate that some Botai horses were bridled, perhaps ridden. Organic residue analysis, using δ13C and δD values of fatty acids, reveals processing of mare's milk and carcass products in ceramics, indicating a developed domestic economy encompassing secondary products.

The domestication of the horse is associated with the spread of Indo-European languages and culture, bronze metallurgy, and specialized forms of warfare (13). Genetic studies of modern domestic horse breeds (Equus caballus) (4, 5) imply either multiple domestication events or domestic stallions from a single original lineage being bred with captured local juvenile wild mares (6, 7), but fail to clearly identify when and where horse domestication first took place. A prime candidate for this locus is the Eurasian steppe, specifically the Botai culture, northern Kazakhstan, in the mid–fourth millennium B.C.E., where faunal assemblages consist almost entirely of horse remains (1, 69). The case for horse herding within the Botai culture includes a semi-sedentary settlement structure, incompatible with hunting mobile wild herds, and skeletal element abundances lacking differential transport patterns associated with large quarries. Tools that were probably used for hide working and producing leather straps predominate over projectile points and other hunting equipment. However, age structures within horse herds at Botai do not clearly indicate a husbanded rather than hunted population (6, 7, 9, 10). Indirect evidence for domesticated horses is strongly suggestive but inconclusive. Here, we discuss three new lines of direct evidence to confirm early domestication of horses in the Botai culture.

Horse metapodia are useful in archaeozoological metrical analyses because of their load-bearing function and proclivity to undergo morphological changes relating to breed and differing physical activities. Specimens were selected from four sites in northern/central Kazakhstan, including Botai (during the 2005–2006 seasons) (11); the Tersek culture sites of Kozhai and Kumkeshu [supporting online material (SOM)] (12); and the large settlement of Kent (13), dating to the late Bronze Age (circa 1300 to 900 B.C.E.), by which time the horses are clearly domestic. Twelve key measurements were taken (14), and principal components analysis (PCA) revealed patterning between sites, with the loading plot indicating that most differentiation related to the ratio of greatest length (GL) against four key width measurements. The ratios between GL and greatest breadth of proximal epiphysis (Bp), smallest width of diaphysis (SD), smallest depth of diaphysis (SDD), and greatest breadth of distal epiphysis (Bd) are indices of general limb slenderness, rather than overall size, and have been used in differentiating equid species (15). All four ratios show exactly the same pattern. Specimens from the two Tersek sites, Kozhai and Kumkeshu, show considerable similarity, whereas the domestic horses from Kent are appreciably more slender. The Botai horses are also significantly more slender than those of the Tersek sites, with the distribution of ratios at Kent and Botai being very similar (no significant difference at 95% confidence interval in a Student's t test of difference between means of the four ratios). However, the Botai specimens showed significant differences in the means of the SD/GL, SDD/GL, and Bd/GL ratios of Kumkeshu and Kozhai at high confidence levels (well above 95%), whereas the Bp/GL test fell just short of a 90% confidence level. Figure 1 shows a scatter plot of two of the mean ratios (SD/GL and Bd/GL) for the sites discussed above and two other published populations: Kuznetsk (16) late Pleistocene horses (deriving from Palaeolithic sites in Novokuznetsk, southern Siberia; being the geographically closest wild horses for which the appropriate measurements are available) and modern Mongolian domestic horses (17). The Botai horses cluster very closely with the Bronze Age domestic horses from Kent and modern Mongolian domestic horses. The Kuznetsk Paleolithic horses appear to be much less slender, and the Tersek population displays intermediate morphology. The domestic populations are clearly more slender and, most significantly, the Botai horses plot with the modern Mongolian and Bronze Age domestic specimens, providing evidence that the Botai horses were domesticated.

Fig. 1.

Scatter plot of mean ratios of measurements Bd/GL and SD/GL on horse metacarpals from different ancient and modern populations. Botai horses plot in association with other ancient and modern domestic populations. Bars are mean ± interquartile ranges (not available from published data for Kuznetsk or Mongolian populations).

We examined evidence for bitting damage resulting from harnessing with a bridle or similar restraint (this may refer to a range of possible mouthpieces, including leather thong bridles), in which damage to the skeletal tissues of the mouth occurs when a horse is ridden or driven with a bit/bridle. A macroscopic method that quantifies bitting damage to the mesial or anterior edge of mandibular second premolars (P2s) was applied (18). When a horse is bitted, the bit is placed in the mouth on the mandibular diastema (the bridge of bone between the anterior and cheek teeth), where it can come into contact with the mesial edge of P2s and cause a recognizable vertical strip of wear through the cementum to expose the enamel. More severe wear exposes dentine below the enamel. However, not all enamel exposure on the anterior border of P2sis due to bitting, and other forms of damage can occur, such as dietary wear, so interpretations must consider the size and shape of the area exposed. The bit also comes into contact with the upper surface of the diastema, which can lead to periostitis at this site (19), and repetitive contact can result in the deposition of pathological new bone or destruction of bone (18). The criteria used in this investigation are based on studies of modern animals (bitted and unbitted) with known life histories (18). From the 2005–2006 Botai excavations, 15 P2s (table S1) displaying full occlusal wear (that is, in excess of 4 years old) and mandibular diastemata were examined. Of nine measurable P2s, specimen no. 7 exhibited changes to the anterior edge that could be unambiguously attributed to bitting damage, as evidenced by wear that penetrated through the enamel to expose the dentine (Fig. 2). Mammalian tooth enamel is very hard (about 300 to 400 Vickers hardness number), and although enamel exposure occurs in both bitted and unbitted equids (cementum can also be removed through dietary wear), dentine exposure in this area occurs only in bitted/bridled animals (18). Two further P2s exhibited bands of enamel exposure that were possibly caused by bit wear, but the form of this is not a clear parallel-sided band, so it is not unequivocal. Application of the mandibular diastema scoring system (18) identified four cases of new bone formation at a level indicating bitting/bridling. The clearly bitted P2 (no. 7) was dated by accelerator mass spectrometry to 4658 ± 33 years before the present (3521 to 3363 calendar years B.C.E., 94.6% probability; Oxford Radiocarbon Accelerator Unit reference OxA-18383), which is consistent with the Botai culture. Thus, 5 out of 15 mandibles studied provided evidence of bitting damage.

Fig. 2.

A Botai stallion's lower second premolar (mesial edge), displaying a clear parallel-sided band of bit wear that penetrates through the cementum and enamel. This morphology and depth of wear occur only in bridled animals [figures 2 and 4 of (18)].

The traditional economies of modern Kazakhstan exploit horses for both meat and milk. Degraded animal fat survives in archaeological pottery (20), and its sources can be classified on the basis of the δ13C values of the major n-alkanoic acids, palmitic (C16:0) and stearic (C18:0) acid, which allows nonruminant and ruminant carcass and ruminant dairy fats to be distinguished (21, 22). Ruminant dairying developed relatively quickly after the domestication of cattle, sheep, and goats (23); direct identification of mares' milk in pottery vessels would be clear evidence of horse domestication. Sampling of modern animal fats, including equine adipose and milk fats from Kazakh animals fed on the natural steppe vegetation, was undertaken. Although horse adipose and milk fats were resolved from the fats of ruminant animals, their δ13C values overlap (Fig. 3A); hence, although equine fats can be detected with this approach, equine milk cannot be unambiguously identified.

Fig. 3.

Scatter plots of δ13C and δD values of the C18:0 and C16:0 fatty acid animal fats of modern reference fats (A and B, respectively) and of organic residues from archaeological potsherds (C and D, respectively) from Kazakhstan. (C) shows the δ13C values of C18:0 and C16:0 fatty acids of 50 analyzable lipid residues from 89 Botai potsherds sampled, and (D) shows the δD values of the C18:0 and C16:0 fatty acids of the residues from potsherds assigned as equine fats. All confidence ellipses are mean ± SD and correspond to the values exhibited by modern reference fats. The residues highlighted in red correspond to archaeological equine milk fats.

To achieve separation, we used compound-specific deuterium isotope (δD) analysis of the major n-alkanoic acids, exploiting the phenomenon that in midcontinental regions, such as the Eurasian steppe, the δD values of summer and winter precipitation consistently differ by >100 per mil (24). Tissue lipid integrates both the water and dietary deuterium signal (25), hence their adipose fat integrates the annual δD signal. However, summer milk fat records only the summer δD signal; thus, the δD values of fatty acids in summer milk and adipose fats will differ because of the large difference in the δD values of summer versus mean annual precipitation. Figure 3B confirms that the δD values of the modern reference horse fats exhibit the predicted difference between adipose and summer milk fats.

The δ13C values from the major fatty acid components of the Botai cooking vessels confirm the preponderance of horse fat residues (Fig. 3C), mirroring the dominance of horse bones at the site. A few residues fall into the ruminant reference distribution, which may well indicate the presence of small numbers of hunted cervids or bovids. Most significantly, the δD values show two distinct clusters. The red points in Fig. 3, C and D, correspond to the respective δ13C and δD values of the same five potsherds. All the δD values of the fatty acid components of these residues exhibit significantly elevated δD values. They very likely derive from mare's milk because of their relative displacement from the major cluster of carcass fats (Fig. 3D and SOM). The relatively higher δD values in the archaeological fats are consistent with increased aridity during this period of prehistory (26, 27).

Although existing archaeological evidence for horse domestication at Botai is inconclusive (10), our new skeletal evidence, based on metacarpal metrics, supports the presence of a proportion of domesticated horses in the Botai herds. Moreover, our bitting damage evidence indicates the use of bridles to control working animals and supports assertions that finds of leather thong–producing tools are consistent with horse domestication (6, 7). Finally, evidence for extensive horse carcass product processing in pottery vessels provides direct evidence for their exploitation as a dietary staple. The demonstration of mares' milk processing confirms that at least some of the mares at Botai were domesticated. The fact that horse milking existed in a region remote from the locus of ruminant domestication in the “Fertile Crescent” and in an area seemingly devoid of domestic ruminants indicates that the evolution of strategies for exploiting animals for their milk was not contingent on the adoption of the conventional “agricultural package,” as it appears to have developed independently in the Botai region.

Supporting Online Material

SOM Text

Table S1


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article