Introduction

During the 6th millennium BCE, continental Europe was marked by a fundamental innovation in the development of the producing mode of subsistence. Based on cereal farming and livestock breeding, this “Neolithic Revolution” (Childe, 1936) was followed by a series of important technical innovations, such as copper metallurgy and the introduction of animal traction. The latter was part of what Andrew Sherratt has described as the “Secondary Products Revolution”, i.e., the exploitation of “renewable“ animal resources (physical strength, milk, wool and manure) that did not involve killing the animal (Sherratt, 1981, 1983). From this perspective, the “animal traction complex” is a late phenomenon directly linked to socio-economic changes that occurred in Europe during the 4th and 3rd millennia BCE. Despite the criticism it has provoked (Vosteen, 1996), this model continues to enliven discussions on the “Neolithic Revolution”, the subsequent innovations and their social consequences (Bogaard, 2004; Greenfield, 2010; Gaastra et al., 2018; Kamjan et al., 2022).

Archaeological data from around Europe such as parietal engravings depicting harnessed ards and carts, bone pathologies linked to harnessing or repeated pulling, and the discovery of objects such as wheels, travois, yokes, ards or archaeological features that like plough marks, presuppose the use of animal traction (Sherratt et al., 2006).

Plough marks are the most tangible, widespread and convincing evidence. They consist of linear depressions filled with sediment of a different texture and colour than that of the surrounding deposits. Such marks can be followed over dozens of meters to form parallel or criss-crossing networks. They imply the use of a specific tool, the ard and its traction by a powerful animal such as an ox (Thrane, 1989).

The discovery of prehistoric plough marks and how archaeologists interpret them as an indicator of animal traction has come up against several major obstacles. The furrows are shallow and often erased by erosion or modern farming practices. They are therefore an extremely fleeting type of evidence and are only observed if they were rapidly covered by sediment to offer sufficient contrast between the underlying ground and the fill of the furrows (see Vanzetti et al., 2019). In addition, these remains are usually very difficult to date. The elements they contain (archaeological finds or organic materials that can be radiometrically dated) are unlikely to be contemporary with the furrows. The only reliable dating comes from their stratigraphic position. They are logically older than the sediments that cover them or the pits that cut them. It is therefore not surprising that discoveries of proven Neolithic plough marks are a relatively rare phenomenon (Fig. 1 and Table 1).

Fig. 1: Map showing the location of the Anciens Arsenaux site (Sion, canton of Valais, Switzerland; yellow dot) and European sites with traces of ploughing dating from before 2000 cal BC (red dots).
figure 1

Detailed data in Table 1. Map base: Natural Earth.

Full size image
Table 1 List of European sites with ard tracks dated before 2000 BCE.
Full size table

Evidence for the use of ards is best documented in northern Europe, especially in Denmark and northern Germany, where the earliest features date to the first half of the 4th millennium BCE (Thrane, 1989; Tegtmeier, 1993; Fries, 1994; Andersen, 2000; Louwe-Kooijmans, 2006; Mischka, 2011; Sørensen and Karg, 2014; Gron and Sørensen, 2018). The Mont Bégo and Val Camonica rock engravings in Southern France and Italy that date to the 3rd millennium BCE illustrate ards pulled by a pair of oxen (Forni, 1998; Huet, 2017). Earlier evidence of soil tillage has been attested to around 4300–4000 BCE in the wetlands of the Swifterbant region of the Netherlands, but these marks appear to have been made using a hoe-type tool with no recourse to animal traction (Huisman and Raemaekers, 2014).

Earlier evidence of the use of plough-like tools comes from the Alpine arc. In Saint-Martin-de-Corléans, Italy ard tracks and bovine hoof-prints predate a series of storage pits dated through some 20 radiocarbon analyses to around 4300–4000 BCE (De Gattis et al., 2018). Excavations on the northern slopes of the Alps at the Welschdörfli site in Chur in Switzerland have brought to light ard furrows in a zone between two occupation levels dated to the first half of the 4th millennium BCE (Rageth, 1998). Although seemingly predating those of northern Europe, the plough marks in alpine regions lack a precise chronological framework and their exact age remains unknown.

The question of the appearance of ploughing techniques is even more important as it is supposed to play a central role in the increasing agricultural output, wealth inequality and social stratification (see Bogaard, 2004 with further literature; Bogaard et al., 2019).

In this paper, we aim to present the results of recent research carried out in the heart of the Alps, in the upper Rhône valley in Sion, Valais, Switzerland, where a robust chronological framework for a series of ard tracks observed in Neolithic levels has been defined.

The Anciens Arsenaux site and its chronological framework

The Anciens Arsenaux site is located in the town of Sion (Canton of Valais, Switzerland), on the alluvial cone of the Sionne, an Alpine torrent that flows through the town and into the Rhône. The site, which extends over 800 m2, was discovered before the construction of an underground silo for the Valais Cantonal Archives and excavated in 2017, revealing alternating human occupation levels and alluvial deposits some ten meters thick. These documented occupation levels span most of the Neolithic period, from around 5200 to 3500 cal BC.

The earliest occupation (ensemble N1; Fig. 2a, right) is an early Neolithic settlement, made up of post-holes and hearths. The pottery, sickle blades, millstones, cereal seeds (wheat and barley; Triticum/Hordeum sp.) and domestic fauna (beef, goat and pig) indicate a mature Neolithic economy. Ten radiocarbon dates (Table 2) date this settlement between 5244 and 4914 cal BC (Fig. 2a, left; see below). These dates are consistent with the pottery (Fig. 2b) that is characteristic of the early phase of the Vasi a Bocca Quadrata, dated in the Po plain and Liguria (Italy) to around 5100–4900 cal BC (Del Lucchese and Starnini, 2021).

Fig. 2: Stratigraphy and chronology of Sion–Anciens Arsenaux.
figure 2

a Chronological summary of the Ensembles N1, AG1 and N2 (results of the Bayesian modelling). The red density plot shows the Ensemble AG1 comprising the ard marks. The full model is based on 30 radiocarbon dates (see Tables 2 and 3). b Photogrammetry of the lower part of the stratigraphy of the Anciens Arsenaux site, with Ensembles N1, AG1 and N2. c Selection of ceramics from Ensemble N1. Nos. 2–3 are fragments of a hollow-bottomed vase characteristic of the early phase of the Vasi a Bocca Quadrata culture in the Po plain, dated to the beginning of the fifth millennium BCE. d Selection of ceramics from Ensemble N2. No. 5 shows decorations typical of the Planig Friedberg group (about 4600 BCE), mainly found in the Rhine basin. Photogrammetry: ARIA SA; photographs and drawings of the sherds: S. van Willigen, InSitu SA.

Full size image
Table 2 List of radiocarbon dates for the Neolithic sequence at the Anciens Arsenaux site.
Full size table

This first settlement phase (N1) is covered by humus soil (ensemble AG1; Fig. 2a, right), sealed by occupation levels and covered by sand and gravel from the overflow of the nearby torrent. At different locations in the excavation, groups of parallel furrows filled with sand and gravel extending over an area of some 30 square metres were observed (Fig. 3), as well as hoof prints left by domestic cattle and goats in a ditch where whitish clays have drained off (Fig. 4). Without any reliable information on the taphonomic processes that led to their deposit, we decided not to date the organic elements contained in the AG1 accumulation. However, Bayesian modelling including the 30 dates from the Anciens Arsenaux stratigraphic sequence dates the furrows of the AG1 accumulation between 5116 and 4708 cal BC (Fig. 2a, left; Tables 2 and 3).

Fig. 3: The ploughmarks groups 364, 65, 500 and 499 at the Anciens Arsenaux excavations (ensemble AG1).
figure 3

a excavation plan; b ploughmarks (group 499 in a) during excavation; c ploughmarks (group 499 in a) after excavation; d excavating micromorphological block EM97 through one of the ploughmark grooves in group 499 (see Fig. 5a for its location), with the analysed thin section shown in red. Images: ARIA SA.

Full size image
Fig. 4: Hoofprints at Sion–Anciens Arsenaux.
figure 4

Goat and domestic cattle hoofprints in depression 365 at the surface of AG1 (Photographs: ARIA SA).

Full size image
Table 3 List of radiocarbon dates for the Neolithic sequence at the Anciens Arsenaux site with detailed results of calibration and Bayesian modelling (Oxcal 4.4).
Full size table

A new settlement (ensemble N2; Fig. 2a, right) with several post buildings is separated from level AG1 by alluvial deposits. Six modelled radiocarbon dates place this occupation in the 4836–4527 cal BC interval (Fig. 2a, left). The archaeological material from this level including a Planig-Friedberg type sherd, a cultural group located in the Rhine basin dated to between 4690 and 4565 cal BC (Denaire et al., 2017), corroborate these dates.

Plough marks of the ensemble AG1

Micromorphological analysis

As the furrows in AG1 correspond in all respects to features generally interpreted as plough tracks (Thrane, 1989; Tegtmeier, 1993; Andersen, 2000; Deák et al., 2017; Rentzel and Guélat, in press), block samples were taken during excavation for micromorphological analysis (Figs. 3d and 5). This approach makes it possible to identify disturbances in the soil horizons caused by tillage (Gebhardt, 1995; Lewis, 2012; Deák et al., 2017). The technique involves examining sediments hardened with synthetic resin under a microscope. It has been successfully applied to archaeological sites for several decades (Courty et al., 1989; Gebhardt, 1995; Lewis, 2012; Deák et al., 2017). Micromorphological analysis was carried out on a 25 × 15 cm sample taken across one of the furrows observed in the AG1 horizon and comprising one of the linear tracks recognized during excavation (sample EM97-1; Fig. 3d). The sample was oven-dried and then impregnated several times with epoxy resin. After polymerization of the resin, the sample was cut with a large-plate diamond saw to produce five thin 42 × 62 mm sections. Micromorphological analysis of the thin slides was carried out using a binocular magnifier and a polarizing microscope (PPL: plane polarized light; XPL: crossed polarizers). After an initial microscopy, the sediments were first correlated, as far as possible, with the stratigraphic documentation from the excavation. Next, the individual micro-layers (mc) were described according to method-specific criteria (Bullock et al., 1985; Stoops et al., 2010).

Fig. 5: Micromorphological analysis results.
figure 5

a View of the sawn face of the sample from which five thin sections were made. The microstratigraphy consists of four units (mc. 1 to mc. 4). In the ploughed horizon (mc. 3), at least three generations of furrow bottoms are recognized by analysis (I–III). At the top, the last generation is filled in by light-grey alluvial sediment (mc. 4) and corresponds to the lineations observed during excavation. b Microscopic view of the transition level (mc. 2). Fragmented and punched fine gravels in situ indicate an implementation with a rudimentary tool. c At the base of the ploughed soil (mc. 3), gravels are bedded according to the lower contact. They correspond to the bottom of ploughing furrows. Bone fragments are probably included in the soil as fertilizer. d View of the contact between, at the base, the ploughed horizon with humic matrix (mc. 3) and, at the top, the carbonated fill (mc. 4) of the upper furrows (III). Gravel layering is also found at the boundary. b, c PPL, d XPL.

Full size image

The parallel furrows uncovered in AG1 have a “U-shaped” profile and measure a maximum of 3.5 m in length, with a width varying between 3 and 7 cm; the spacing between the features varies from 10 to 20 cm.

Thin section analysis has identified a microstratigraphy composed of four units (Fig. 5a, mc. 1–mc. 4). At the base, carbonate alluvia, a low-grade soil (Fig. 5b, mc. 2) cover fine torrential deposits (mc.1). Certain micromorphological features, such as the in situ verticalisation or fragmentation of elements, suggest that the soil was worked using a rudimentary tool, such as a hoe (Lewis, 2012; Gebhardt and Langohr, 2015). Traces of frost detected in the base gravels (mc. 1) could be the result of soil denudation at this stage (Curdy and Guélat, 2011). The sharp, undulating upper boundary is marked by a packed silty level. Above this, the matrix becomes homogeneous and contains degraded organic matter (Fig. 5c, mc. 3). This humus-bearing soil, eroded at the top, also contains anthropogenic components such as bone fragments and charcoal, perhaps added as a fertilizer (Lewis, 2012; Devos et al., 2013). This same soil contains at least three generations of splayed «U»-shaped features, 10 cm wide and up to 3 cm high, also individualized by layered gravel at their lower limit and corresponding to the bottom of the furrows (Fig. 5c). These features are located above the lower horizon contact, also marked by the compaction of the underlying sediment (mc. 2), the latter having also undergone internal silting.

Similar microscopic features are the main stigma that shows the use of implements (Deák et al., 2017). The U-shaped tracks can thus be interpreted as the bottoms of ard furrows. An analogy with Iron Age plough marks identified at the nearby site of Gamsen (Valais, Switzerland), also located on a torrential cone of the Rhône valley in a morphosedimentary context comparable to that of the Anciens Arsenaux corroborates this interpretation (Rentzel and Guélat, in press). At the top of the sample, alluvial sediments (Fig. 5d, mc. 4) fill the most superficial furrows that were observed during the excavation.

Micromorphological analysis therefore indicates that the furrows were formed by tillage, suggesting the repeated passage of an ard after the initial substrate preparation. The three generations of furrow bottoms show that this process was repeated several times. The anthropogenic input of nutrients (charcoal and bone fragments) into the ploughed soil confirms agricultural practices on the site.

Despite the small size of the ploughing area that has been preserved, the regularity and continuity of the traces, the compactness of the worked sediment and the similarity with the marks identified at Gamsen suggest that animal traction was used.

Dating the plough marks of the ensemble AG1: Bayesian modelling of radiocarbon dates

Stratigraphic information

The chronological model developed here relies on a series of 30 radiocarbon dates (Tables 2, 3) and the stratigraphic succession of eight ensembles of archaeological contexts. The first ensemble (N1inf), located at the base of the stratigraphic sequence, yielded three burnt tree stumps, each sampled and dated. These three dates represent a terminus post quem for the site’s Neolithic sequence, itself made up of seven superimposed stratigraphic ensembles (N1–N6). Each of these sets has been radiocarbon dated: N1 (10 dates), N2 (6 dates), N3 (5 dates), N4 (2 dates), N5 (3 dates) and N6 (1 date). The ensemble AG1, characterized by the linear structures that are the subject of this article, yielded no diagnostic archaeological material or carpological or faunal remains suitable for radiocarbon dating. However, the stratigraphic insertion of AG1 is clearly established (AG1 is posterior to N1 and anterior to N2). This stratigraphic sequence enabled us to construct a model of strict succession of eight phases, incorporating chronological information derived from the radiocarbon dates of each assemblage: N1inf < N1 < AG1 < N2 < N3 < N4 < N5 < N6. Due to the scarcity of short-lived materials on the site, charcoal was favoured. Although an old wood effect in the sequence cannot be ruled out, its impact may be limited by the number of dates and the modelling.

Bayesian modelling

The stratigraphic diagram allowed designing a model of eight sequential phases using Oxcal software. The model includes start and end boundaries for each of the eight phases. The reason for modelling a sequence of phases was that, although the upper and lower boundaries of each ensemble were clearly defined, it was not possible to establish all the stratigraphic links between the different dated events within each ensemble and therefore to model these links within each ensemble as a sequence. Only the succession of three successive ploughing events identified in the micromorphological analysis of AG1 (Fig. 5a) could be modelled as a sequence (see below).

Calculations were performed using OxCal software (version 4.4.4; Bronk Ramsey, 2009a) and the IntCal20 calibration curve (Reimer et al., 2020; Table 3). They include a full outlier model (Table 4; Bronk Ramsey, 2009b). The chronological information corresponding to each phase was summarized using a Kernel density evaluation (KDE) with the KDE plot command (Bronk Ramsey, 2017). As no radiocarbon dating was available for ensemble AG1, the date of each of the three successive ploughing generations associated with this ensemble was estimated using Oxcal’s Date command, with the succession of the three dates modelled as a sequence (see code in Supplementary Information). These three estimates do not provide any new chronological information, but they do enable us to estimate the ploughing date and calculate a KDE for the ensemble AG1. The model thus provides a complete overview of the entire sequence.

Table 4 Detailed results of the outlier analysis of radiocarbon dates from the Neolithic sequence at the Anciens Arsenaux site (Oxcal 4.4).
Full size table

Modelling results

The results of Bayesian modelling generally show a very good correlation between radiocarbon measurements and the model, with an agreement index for the model of 86.7% (AModel) and an overall agreement index of 82.3% (AOverall) (Table 3). Only one modelled date has an agreement index below 60%, the threshold below which the date in question should be reconsidered. Outlier analysis shows that the vast majority of dates have an a posteriori probability of being an outlier of less than 7%, most of them being below the 5% threshold (Table 4). Only one of the three dates in ensemble N5 (Poz-112400) is questionable since it has a posterior probability of being an outlier of 50%. As ensemble N5 is at the top of the stratigraphic sequence, this problem only marginally affects the dating of assemblage AG1.

The modelled dates (95.4%) (Table 3) indicate that ensemble N1, corresponding to the early Neolithic, is set between a Boundary Start of 5244–5045 cal BC and a Boundary End of 5203–4914 cal BC. The estimated dates correspond to the three successive ploughing episodes identified in ensemble AG1 (AG1_Ev1 to AG1_Ev3) cover intervals between 5116 and 4708 cal BC. The N2 ensemble ranges from a lower bound of 4836–4625 cal BC and an upper limit of 4708–4527 cal BC.

Early emergence of animal traction in Europe

The plough marks at the Anciens Arsenaux site appear within a dilated stratigraphic sequence and are therefore reliably dated by both radiocarbon analysis and pottery typology to between 5100 and 4700 cal BC. They predate by around a millennium the earliest traces of ploughing in Denmark and northern Germany appear around 3700 BCE (Sørensen and Karg, 2014). The Anciens Arsenaux site therefore documents the early use of animal power in the Alpine arc, already recorded in the area in more recent or contemporary contexts at Aosta-Saint-Martin-de-Corléans, Italy (before 4300–4000 cal BC) and at Chur-Welschdörfli, Switzerland (before 3500 cal BC).

The age and location of these tracks tend to corroborate the fact that evidence of Neolithic ploughing is only found on the geographical margins of the two major drifts of Neolithisation in Europe: to the north of the Danubian drift (Denmark, Northern Germany, Netherlands) and to the north of the Mediterranean drift (major Alpine valleys) (Fig. 1 and Table 1). This observation can be interpreted in two different ways:

- The beginnings of ploughing are the result of a late phenomenon of adaptation at a time when Neolithisation was affecting areas less suitable for agriculture than the great European plains.

- The map of known Neolithic ard tracks does not show the real distribution of the use of this tool during the Neolithic. It indicates the differential preservation of plough marks between the lowland areas where the intensive farming practised after the Neolithic has all but erased these features and specific contexts such as dunes, tumuli, and volcanic ash layers (see the case of the Bronze Age fields in Campania; Saccoccio et al., 2013; Vanzetti et al., 2019) or alluvial deposits with a high sedimentation rate that has led to their preservation.

New discoveries from the Anciens Arsenaux site contradict the first hypothesis of the development of animal traction when reaching new environments during a late phase of Neolithisation. Traces of ard tracks discovered at the Anciens Arsenaux site, dated between 5116 and 4708 cal BC, are the oldest known in Western Europe and show that the use of animal traction appeared very early in the Alpine Arc, in a chronological interval immediately after the first appearance of a production economy in this region.

As already suggested (Lüning, 1980; Helmer et al., 2018), these new agricultural techniques could well be an integral part of the original Neolithic European package. This model is consistent with archaeozoological studies that suggest the early use of animal power, at least occasionally or not very intensively in several areas in southern Europe. The evidence dates to the middle of the 9th millennium BCE in northern Mesopotamia (Helmer et al., 2018), the beginning of the 7th millennium BCE in Anatolia (Kamjan et al., 2022), the 7th millennium BCE in Crete (Isaakidou, 2006), between the end of the 7th millennium and the middle of the 5th millennium BCE in the western Balkans (Gaastra et al., 2018) and the end of the 6th millennium BCE in the western Mediterranean (Helmer et al., 2018).

The absence of ard tracks on the great European plains, which are among the first areas to see early Neolithic farming, is therefore likely to be linked to taphonomic conditions unfavourable to the preservation of these fleeting vestiges. The great Alpine valleys, on the other hand, where human settlements are often located on alluvial fans, have the capacity to preserve such traces because they were protected by sedimentation. Moreover, these traces are found in dilated stratigraphic sequences rich in organic remains that guarantee their correct dating. Future research in comparable environments is likely to provide new, reliable and precise data on the introduction of animal traction in Western Europe.

Conclusion

Our research has provided a solid chronological framework for the earliest known plough marks in Europe, dated between 5100 and 4700 BCE. These remains demonstrate that the use of animal power appeared quite soon after the first evidence of a production economy in the Alps. The new data indicate that the use of animal traction did not develop during a late phase of the Neolithic in Europe but was probably an integral part of the initial processes of the continent’s Neolithisation.

Animal traction is an important innovation that may have had considerable implications for economic and social development during the Neolithic period, mainly in terms of increased output and subsequent wealth inequality. Early emergence of ploughing in Europe, as suggested by new discoveries in the major Alpine valleys, should at least prompt a reconsideration of certain points in A. Sherratt’s “Secondary Products Revolution“ model, in particular the question of farming practices and social organisation during the early Neolithic.