On the upper terrace, generally between 1.50m and 2m of soil is missing. BU1 includes a layer of the 1st century BC with large blocks aligned and containing armaments. BU2 is a reference for the final Bronze IIIb and the first Iron Age. BU10 comprises a layer of the same nature as that of BU1, topped with a level of destruction of the same period. On the lower terrace, BU3 corresponds to a garden layout in the 16th century. BU4, surmounted by old excavations, has at its base a surface of Gaullic circulation with a hollow structure. BU5 to BU12 are piles of excavated material accumulated by the ancient excavations. |
Surveys of the current topography, set against the analysis of the documentation (in particular the information on the ancient excavations) plus the geological and hydro-geological studies, made it possible to come close to defining the outline of the topography of the sector of the Loulié spring in the 1st century BC. Since the beginning of the Holocene a mass of travertine has gradually formed at the foot of the cliff of Puy d'Issolud, caused by the presence of a permanent, abundant spring at this location. This massif then experienced a succession of phases of erosion and accretion, caused by man-made impact on the surrounding environment. The mass, existed in the 1st century BC in the form of a shelf at the foot of the cliff. The shelf contained a series of natural basins filled with water, spread over an area of several hundred square meters, developed by man over time A cascading front allowed the basins to overflow. Archeological and geological studies have shown that this front advanced much more towards the valley than its present geometry suggests, since it is today greatly reduced by quarries and by both medieval and modern agricultural practices, by earthworks generated by the excavations of the nineteenth and early twentieth century and also by the mechanisms of erosion specific to this type of geological formation. The impact of erosion must be taken into account all the more because nothing in the first stratigraphic studies of the mass suggests that the phenomenon of accretion took place after the period of the second Iron Age. |
On the south-east side, at the level of a fracture, was a large deposit of fine clay. The archaeological remains here consist of three coins from the early 20th century and two modern shoe nails with round stems. Up to the last step, the infill was composed of large and medium blocks of limestone mixed with a clayey earth, containing a few small pieces of quartz gravel. Some contemporary furniture items have also been discovered there.
Basically, the infill consisted of a very clayey earth containing some small quartz gravels and very small calcareous elements. No archaeological element was found there. The notes left by J.-B. Cessac concerning his excavation of the sector confirm that he found no remnant of Iron Age or any other period in the basin itself. The few weapons he mentions in this sector come from the slopes on the periphery of this basin. All these elements lead to the conclusion that this basin is not a sustainable source and especially that it was probably not known in the Gallic period. The intersection of these two approaches makes it possible to affirm today that the "Cessac basin" can in no way be the water source evoked by Caesar's text and that, given the geometry of the travertine massif, the agger built by the Roman troops could not be as high on the slope of the site as Napoleon III thought. This agger must therefore be sought much lower. |
The main results in this area come from the excavation of the BU10 mound on a conserved surface of about 16 m2, which revealed a layer of destruction of a burned Gallic construction. The presence of reddened earth, charred logs and Roman armament made it possible to date this structure.
The research has highlighted three main steps. First of all, a 6.5 m wide excavation was made in the final Bronze layers, up to a natural scree on the slope. The old excavations have amputated this construction on its eastern and western flanks. The soil at the bottom of this excavation was very irregular; large blocks of natural scree are apparent. In its northern part, about 3 m wide, the ground is pretty much horizontal. To the south and south-east, it has a southwest-facing slope of about 20 °, while to the northwest it has a strong east-west slope. The numerous bronze fragments of the final Bronze IIIb present at the base of this construction are mostly crushed by trampling, thus confirming the existence of a floor of circulation; this one is strongly reddened under the action of fire; five ancient arrow heads had crossed the overlying layer to fuse into the Late Bronze layer. This reddened ground was covered with a layer approximately 10cm thick containing numerous objects, all of which have undergone the action of fire (Fig. 10):
|
On the surface of this layer were numerous fragments of charred oak logs. In their positions, the very compact and strongly rubefied soil was carbonised in places. The 14th century datings made on five of these logs yielded the following chronological calibrated ranges: -103 to +82; -86 to +76; -356 to -2; -85 to +67; -43 to +78. Two archeomagnetic analyses of the reddened soil yielded the following age ranges: -100 to +10; -100 to -5.
This layer was covered by a scree of thickness 0.35 to 0.60 m, completely reddened. About 4 m wide, it was made up of terracotta, bricks (originally moulded mud blocks) fired during the fire (2) and travertine rubble altered or converted to lime by the action of fire (Fig. 11). In its southern part, 1.60 m wide, the scree was mainly composed of reddened limestone, agglomerated by heat. But to the east, at the base of this area of stone, there were fragments of bricks (originally blocks of raw earth) and travertine blocks altered or converted into lime by the action of fire. Above this first scree, probably resulting from the collapse of an elevated structure, was another reddened scree, with a thickness of 0.15 to 0.25 m, consisting of medium and large limestone elements also altered by the action of fire. This layer is absent to the south. The archaeological elements of the upper scree consist of Dressel 1B amphora fragments, the bottom of the handle and the foot of a Lamboglia 2 amphora, small fragments of Gallic ceramics, and a tapered broken tang of rectangular section, with its ferrule. Almost all the elements that made up the two screes have a general east-west or north-east / south-west dip. Finally, to the south-east, we uncovered an accumulation of unreddened stone blocks that appear to be in place. |
Research in mound BU1 revealed an interrupted soil in the south lying against large blocks. The layer, 3 to 10 cm thick, that covered this ground contained 7 arrowheads and a catapult bolt. An arrow was planted in the ground. This layer contained 13 pieces of charcoal. Pottery was scarce and limited to about fifty small pieces made of native gray or beige clay and 38 Dressel 1 type amphora fragments, including a handle that is attributable to Dressel 1B. This layer was itself covered with a landslide of blocks. |
Work on mound BU4 revealed a small hearth with an undated cultural layer. Under this layer was a landslide of blocks that appeared to come from a collapsed dry stone building, which would have been located further east, or southeast. Under the blocks and above the travertine bedrock, the excavation revealed a cultural layer with a hollow, rounded structure (1.5m diameter to 1.1m depth). In this excavation and on its periphery, have been discovered artifacts attributable to the late Tene culture, without any more precision, in particular fragments of dressor 1 amphora belly and handle, of native gray clay pottery without particular character, and nails. No evidence of fire or armaments was found. |
To the east, the presence of Dressel 1/1B amphora fragments and some Gallic pottery shards practically in contact with the substratum allows us to say that, at least in the first century BC, the landscape of the slopes under the cliffs was totally bare. The steepness of the slopes, of the order of 35° to 40° depending on the sector, prohibits any human occupation. No development has been identified in this area. The careful examination of these few "shreds" of archaeological layers in place and the artefacts they contained confirms two important points. There was indeed an occupation of the second Iron Age centred on the sector of the springs located on the mass of travertine at the foot of the cliffs; it is characterised by a layer and a level strongly burned by a violent fire, and delivering armaments in abundance and limited quantities of various pottery artefacts. These are clearly attributable to the 1st century BC. |
The armaments are in all respects similar in composition to those discovered in the ditches of Alesia or Gergovie and more widely on all sites of battles of the Caesarean period found in Gaul. The catapult bolts and iron nails with their special heads are the main markers. What stands out here, compared to other sites of the same type and of the same period, is the abundance and the concentration of the armaments. Taking old discoveries and recent excavations together, more than 1260 arrowheads and a hundred horseshoes have been unearthed over an area of just 4000m2! The radiocarbon and archaeomagnetic datings are themselves all compatible with an attribution of the whole to the 1st century BC, more precisely to the Tene D2. Moreover, this occupation was obviously of short duration and of a very specific nature, judging by the thin layer containing the artefacts, the absence of significant furniture before or after the current of the building. 1st century and the concentration of objects in the military field. |
It is these underground structures that have attracted the attention of researchers and scholars since the nineteenth century, leading to the almost complete destruction of the archaeological levels in place.
It was therefore essential to resume work in this area as well, both to precisely locate the sections seen by the various past researchers (notably J.-B. Cessac, Napoleon III and A. Laurent-Bruzy), and to try to better understand the network, its purpose and its chronology. The research on these tunnels comprised many activities of different natures: successive soundings, manual clearing when they were found to be filled with earth, a complete topographic survey, a study of the way they were dug and analysis of the geological context. The clearing of sections already identified led to the discovery of new sections. From the stream, downstream of the site and quite low down the slope, a ditch was dug under the agger. In view of the Gauls, the Romans passed underground. The network rises steeply and bifurcates when it reaches the foot of the bed of travertine on the surface of which is the source. 1.5 m wide and 2 m high on average, these vaulted tunnels extend over a recognised length of 111 m (69 m for the southern branch and 42 m for the northern branch). |
From west to east lies the main G8 tunnel, partially discovered by A. Laurent-Bruzy in 1935, which seems to have divided in its upstream part into two longitudinal branches. One, to the south, was found in 1865 by J.-B. Cessac (G6). The other, to the north, includes sections G7, G1, G2, G3. Two upper tunnels G4 (14 m long) and G5 (length 8.40 m) were dug by the Gauls to intercept the Roman sappers (Fig. 16 and 17).
This very complex network of tunnels was the subject of an archaeological survey report by Jean-Claude Bessac (CNRS, UMR 154, Lattes) on digging techniques and strategies. According to this report, no tunnel discovered to date can be associated with stone quarrying, as much because of the very poor quality of the travertine as of the relative narrowness of the tunnels and their vaulted profile. In addition, the impacts of tools found on the walls do not correspond to those found in quarrying. The hypothesis of the capture of a spring by underground pipes to feed a mill or for any activity requiring abundant water cannot be excluded outright. However, no such installation was found downstream. But above all, it is hard to see why, in this hypothesis, two networks of tunnels would have been dug, a rectilinear one (north network) and the other broken (Cessac tunnel) and how to explain the two side tunnels G1 and G5. |
The northern tunnel system did not enable the Roman sappers to find the black marl. Tunnel G3 seems to have been abandoned due to the difficulties of digging in unstable ground composed of soggy clay and sloping scree, but it is more likely to have been because of the Gallic counter tunnelling (tunnel G4). The G5 side tunnel could also be a counter tunnel to intercept the Romans at the Cessac tunnel. The tunnel abutted in gravel formed by repetitive action of frost on the rock and would not have allowed the Gauls to go further. According to J.-Cl. Bessac, the technical arguments in favour of a meeting of two teams of Roman sappers equally apply to a team of Gauls who would have worked in the G4 tunnel to try to stop the progress of the Romans. In this hypothesis, the G4 tunnel would become a counter tunnel and its annex G5 a research tunnel that would have been oriented to the south by the tool-cuts of the Roman sappers of G6 digging about 6 or 7 meters away. But the meeting between Gauls and Romans would have been faster in G3. To this hypothesis, one can consider in the first place Caesar's text in The Gallic Wars (in fact the editor Hirtius) which implies that the Gauls did not understand the Romans' tactics, and surrendered believing themselves abandoned by the gods after the drying up of the spring (7). The Romans often praised the strength and especially the bravery of the Gauls, because defeating them increased their own prestige. But they rarely applied this rule in terms of military intelligence, especially under the pen of Hirtius, who did not actually experience the siege. The affirmation of the Romans would only have served to demonstrate the intellectual inferiority of the Gauls. In the second place, it would also be necessary to explain the digging of the G1 tunnel by the Roman sappers. |
An initial explanation leads one to consider the additional drying up of secondary springs. But we can also consider a faster connection to tunnel G6, on the south side. The objective of the extension of G2 (beyond G1) and G3 tunnels by the Romans, who had certainly realised the low hydrogeological interest of the sector and the bad state of the ground, would have been then purely military - to divert the Gauls' attention and possibly confront them underground to allow time for the team of the south tunnel (G6) to finish their work. In addition, the desire to intercept the G5 sappers who dug 8 meters to the east, is not excluded. The hypothesis of a Gallic counter tunnel should therefore imply, in parallel, skirmishes, or even a fight in the tunnel G2-G3. No archaeological evidence attests to such military engagements (8), but the use of certain military devices, such as the use of smoke in tunnels, would leave almost no trace, especially in these places where the deposits of calcite quickly change the surface appearance of the rock. In favour of this last theory, we know that the use of counter tunneling is attested in antiquity (9). On the other hand, the hypothesis does not require the discovery of a junction between tunnels G6 and G4, since the Gauls were in their camp. They were able to dig a shaft for the descent to start digging the gallery in the area of their spring, even though they had to protect themselves from Roman fire during this operation. The presence in tunnel G4 of holes for hooks to support oil lamps does not seem to be a frequent Roman practice, but we do not know about the Gauls' practices. In Book VII of the The Gallic Wars, on the subject of Avaricum's defense, Caesar notes: "They (the Gauls) were crumbling our earth by digging counter tunnels, all the more knowledgeable in this art, as they have great iron mines and they know and employ all kinds of underground galleries."(10) We cannot therefore question the technical abilities of the Gauls in this area. |
There is no text that specifies the distance of the tower from the spring. It cannot exceed the range of effective shots by catapults, bows and slingshots. It must be far enough away so as not to be vulnerable, especially to flaming projectiles. From the tower alone, the Romans' projectile fire was not sufficient to be decisive. It did, however, greatly hamper the besieged who, exasperated, threw flaming barrels filled with tallow, pitch, and thin slats of wood (B.G., VIII, 42) from the travertine platform against the Roman construction. Hurtling down the slope to the bottom, they set fire to the mantelets (wooden protective shields) and the base. Simultaneously, the Gauls made an attack in force to absorb all the Roman forces, to stop them working to extinguish the fire. Caesar then gave the order that, on all sides, the troops should begin to climb the slopes and shout, as if they intended to cross the Gauls' fortifications (B.G., VIII, 43). The leaders of the Gauls recalled their combatants and placed them all over the walls. The fight having ended, the Roman legionaries were quick to put out the fire. It is surprising that the Roman troops were content to simulate an assault. Could they not have carried it out at this time or even much earlier? No doubt, but here as elsewhere, Caesar, anxious "never to sacrifice his soldiers of elite, rare and precious beings ... prefers a deadly escalation of a few minutes to the interminable fatigues of a blockade" (11). Frontin already explained that Caesar had adopted the medical motto: "Rather the diet than the scalpel" (12). Since the failed assault of Gergovie (B.G. VII, 47 sq.), Caesar, suspicious, halted before the slightest enemy fortification (B.G., VII, 69-70). The agger and the tower, as imagined by Jean-Baptiste Cessac, Napoleon III and Armand Viré, were at the foot of the travertine massif and, according to Etienne Castagné and other authors, on the platform of travertine. |
The tower was considered to be located at a distance of 20 to 30m from the Cessac spring, about 40 to 50m from the cliffs. This does not accord with the motives that led Caesar to use artillery to prevent the besieged Gauls from approaching the spring. The harassment of the Romans made the Gauls' access to the spring perilous. Taking water up onto the oppidum would have required a procession of water carriers from dusk until dawn in order to have the best security. Constructions put in place to protect these activities from attack would seem to be essential. These could have been walls and palisades, behind which the water carriers could shelter. If such a protective arrangement had not been constructed, access to the spring would have been severely disrupted by Roman fire from the tower. And a significant reduction in the water supply would also have shortened the length of the siege. But it was necessary to resort to capturing the spring by subterfuge, to remove access to this essential water supply. This proves that the Gauls in charge of fetching water were able to access the spring, bearing risks of course, but minimal risks. The contrary would have quickly brought an end to the siege. New research has shown that the Gallic basin discovered under Napoleon III was a simple natural cleft in the rock not known to the Gauls. The true basin would have been located about twenty meters further downstream, on the travertine platform, simply at the site of the primitive washhouse. The discoveries made on the south side of the site (in E20) show that the end of the agger, positioned by the former researchers, was located inside the Gauls' defenses. The shooting trials, carried out in 1998 and 2005 at the site of Loulié, using reconstituted weapons, made it possible to note that the tower was entirely in range of the enemy. |
In 2005, surveys conducted downstream of the site by Hubert Camus (of Hypogée) as part of the geological and geomorphological study of the site, led him to consider that the limestone blocks and coarse gravel encountered in the surveys and brought to the surface during the storms of 2001 (?) and intersected in soundings could well represent the remains of the Roman agger or associated military infrastructure. Several characteristics distinguish them from the infra-travertine large block formation: they are not altered, they are packed in laminated clay sediments and not in an alteration clay, they are associated with a terrestrial malacofauna and an outbreak of fire in place dated to the final Tene (Calibrated age 195 BC to 1 BC). A resurgence of water between the stream and the house is in the axis of the main tunnel (Fig. 19). It already existed in the nineteenth century. It lies exactly in the axis of the main branch of the tunnels found by two soundings downstream and upstream of the house. The particularly minute excavation of the Gallic layers was the object, in addition to the classical coordinates, of a precise reading of the orientation and inclination of metallic objects (nail, arrowhead, catapult bolt). It was also noted whether the arrowhead or the catapult bolt was stuck in the ground. That is to say that the projectile had gone through at least 10 cm of the Gallic layer (floor of circulation or wasteland). We used only these objects to determine the origin of the shots. The orientations of the projectiles concerned were reported on the topographical plan of the site at 1/100th (13). |
The results gave as origin: 8 projectiles coming from parcel 176, and 4 projectiles coming from the stream and the house. Eight trajectories are 25 m wide from the edge of the current stream to the house. These results must be interpreted with great caution, because of potential errors: the object could have moved in the ground, its direction could be inaccurate due to its short length, as could its recording on the plan. Logically, but with reserve, we can position the agger and the tower north or north-west of the house. If one positions the tower between the stream and the house, the distance of the tower from the Gallic road would be of the order of 80m, it would overhang it by about 8 meters. If you fix the tower at the level of the road, you go back about 35m. It would overhang the path by about 4m. The limit of the shots would be on a radius of 115m from the tower. The maximum distance must not exceed the range of effective shots of catapults, bows and slingshots. The distance from the Gallic basin would be about 95 m. The inclination of the projectiles gives us an idea of the angle at which they were fired. Assuming that the tower overhung a few meters up the path of the water carriers up to the plateau, the angle of the shots ranged from 8° to 27°. For the two shots close to 50°, the archer sought a maximum range by taking a sighting angle of about 45°. The shooting experimentation carried out in 1998 and 2006 from the cliffs overlooking the Loulié fountain and the proposed site of the Roman tower, provided us with very useful information on range and the zones of insecurity and provided an understanding of the flight qualities of the projectiles and their effectiveness (Fig. 18). |
The work undertaken on the site of the fountain of Loulié over a period of 12 years leads to an updated reading of the question of the siege of Uxellodunum. First of all, the field data are now sufficiently numerous and consistent to confirm whether the interpretations of J.-B. Cessac and, in his wake, of Napoleon III concerning the location of Uxellodunum, were relevant. Field archeology shows us that the Loulié spring was indeed the theatre, around the middle of the 1st century BC, of a violent fight bringing together Roman troops against the Gauls. The abundance of armaments from the Caesarian period is impressive and unique in comparison with all the other battlegrounds of the same period identified in Gaul; this armament is clearly concentrated on the perimeter of the springs which obviously constituted the stake of the combats; the digging of underground tunnels obeys a logic directly related to the plan to dry up the springs; in this sector of Puy d'Issolud, the occupation of the second Iron Age seems tightly limited to the sector of the source and the period of the siege. |
Archeology has thus managed to recover most of the elements described or evoked by the text of Hirtius on the battle around the source of Uxellodunum: an abundant source of water located on the side of an elevated site powerfully defended by its natural characteristics, which was the scene of violent fighting and that the Roman troops dried up by digging underground tunnels. It is too rare for archeology to reveal so many obvious and concordant elements not to highlight it here. It should also be remembered that information from medieval texts and from toponymy are themselves in full agreement with the archaeological data. The aerial works built by the Roman troops to ensure the siege of the source (including agger) described by Hirtius cannot be sought where Napoleon III proposed to locate it. The hypothesis defended by Napoleon was indeed the following: the Gallic basin found in 1865 by J.-B. Cessac would be the source diverted by the Romans. It was allegedly defended mainly from the cliffs by rebels and archers. The Gauls from the "Pas de la Brille" would have descended, to fetch water, by two trails east and north-east of the source. |
Two elements suggest the location of the imposing Roman terracing. The first is the mapping of the located weapons, including arrowheads and catapult bolts found in their primary position. This suggests a focal point from which shots were fired a short distance north of the current house west of the source, more than 50m below the position proposed by Napoleon. The second element is the presence of the main branch of the tunnels, in the form of a trench, which was found by two soundings (downstream and upstream of the house). Downstream of the house, a trial hole revealed a hollow structure with large rock blocks, attributable to final Tene (by the archaeological material) which could be a Roman construction linked to the agger. In another trial hole, at a depth of 1.10m, a large combustion zone containing rusted stones and charcoal was dated by radiocarbon to the final Tene. The excavation of a few shreds of archaeological layers preserved from the picks of the old researchers found the soil strongly reddened by the violent fire described in the nineteenth century. Under the action of fire, raw bricks (from Gallic defensive constructions) have cooked and taken a color ranging from pink to dark red, travertine blocks have been altered to become lime and building stones have been heavily reddened. The battlefield has delivered a large new batch of Gallic and Roman armaments, some Gallic pottery, charred wood pieces and fragments of Italian and Spanish wine amphorae. All the techniques used to date the use of this soil give the same results: the radiocarbon analyses of charred wood logs and paleomagnetic analyses of the sediments affected by the fire as well as the study of the objects, place the event towards the middle of the 1st century BC. The presence on this soil of a large number of oval pebbles weighing 50g to 2kg, from the bed of the Dordogne and therefore inevitably brought by man, suggests that they are projectiles. The smaller ones were able to be thrown with slingshots while the heavier ones could serve as ammunition for the ballistae used by the Roman troops. |
The resumption of the study of the underground tunnels has yielded key information, and has shown that only the Cessac tunnel has resulted in the capture of the water sources that fed the Gallic basin. However, despite the important investment of several years by a team with multiple skills, there are still outstanding questions: Hirtius evokes Uxellodunum as a stronghold of the Cardurques "remarkably defended by nature". However, to date, the mimimal archaeological research carried out on the plateau of Puy d'Issolud has provided little information about the Gallic occupation. The remains from this period are few in comparison to the abundance of those discovered in the sector of the Loulié spring. The plan of the fortifications proposed by Castagné would need to be validated, because the chronological argument proposed for the rampart is not solidly supported. It would also be interesting to look for the circumvallation which Caesar and his troops built around the site at the time of the siege and to look again at the question of the location of the three Roman camps established by Caninius and mentioned by Hirtius. Fabius and Calenus having subsequently brought four and a half legions must necessarily established other camps. These new camps were not built in preparatuon for a Gallic attack coming from outside, but with the single objective of investment. New excavations would verify and date these protohistoric fortifications and identify the nature and importance of occupation of the site at this time. The monographic publication of all the data from the old works and our own research is completed, the European Archaeological Center of Mont Beuvray, under the direction of Vincent Guichard was in charge of editing and publishing. The Uxellodunum Site Management Joint Syndicate (S.M.G.S.U), which was the prime contractor, was responsible for the development of the Loulié site. The purchase of the house below the site was carried out in 2009, the acquisition of the land, the restoration works of the galleries, the earthworks and the modeling of the site were carried out in 2010. |