1From October 2002 to June 2004, 16 wrecks dated to between the first century A.D. and the 18th century were excavated on the ancient right bank of the River Saône in Lyon, France. The Institute for Preventive Archaeological Research (INRAP) at the request of the Ministry of Culture carried out the excavations of this harbour site, located on the future Parc Saint Georges’ car park. Grégoire Ayala was the manager of this huge archaeological intervention and Marc Guyon was responsible for the excavation of the wrecks. Operations took place within a cofferdam following traditional methods of terrestrial archaeology on an area of approximately 4,000 m2, to a depth of 10 m (Ayala, Guyon, Laurent 2005).
2The Lyon Saint-Georges 4 (LSG4) wreck belongs to a set of six Gallo-Roman boats discovered in 2004. A complete architectural study was conducted and published in 2010 (Rieth, Guyon 2010), but lack of time and the impossibility of dismantling prevented an in-depth survey. The three most complete wrecks were removed and stored in a lake near Lyon to await project funding. This occurred in 2013 and would ultimately lead to the launch of the LSG4 project, which was designed to run for several years and involved: the raising of the wreck from the water (as with AR3); the consolidation of the waterlogged wood; restoration of the ensemble; the creation of a suitable presentation environment; and the conduct of a complete archaeological study. This large cargo boat could carry a load of up to 53 tons. A full reconstitution of this boat would reach 28 m in length and nearly 5 m in width, however only 15 m in length are preserved. The other parts of the wreck are still under earth, protected by the houses and streets of Old Lyon.
3Eight weeks were necessary to remove LSG4 from the car park, but five months were needed to conduct a complete dismantling and to prepare it for PEG consolidation once the sections were raised and transported to the ARC-Nucléart laboratory in Grenoble, France.
4The second phase of the conservation campaign began in January 2014. The dismantling took three weeks per section. Elements were stored on shelves by architectural type (flat bottom, sides and ribs) and consolidated in a PEG 2000 solution for at least eight months. The various elements were then freeze-dried using two freeze-dryers. It took between two to six weeks for a cycle and 18 months for the whole shipwreck. Since the end of 2016, it has been stored at the ARC-Nucleart lab, waiting for an exhibition space in the Lugdunum museum to be ready in 2024. Time was thus available for further analysis of the various materials and data recovered leading to new archaeological information.
5Complete coordinated dismantling is a real opportunity for the acquisition of all kinds of archaeological data, only 25% of which is available when the wreck is not dismantled. (This estimate is based upon the notion of a section still on the excavation site and studied over three days without special equipment and only partial dismantling, versus a complete dismantling with the intervention of conservators). The removal of all frames gave access to the flat bottom and led to a completion of the original plan that had been recorded during discovery in 2003. This initial plan had left many questions unanswered concerning the assembly of planks in the same strake, and had provided no view of the underneath. Dismantling allowed among other things the drawing and recording of all the surfaces of the constituent elements and both sides of the flat bottom once cleaned.
Fig. 1: A crack in plank BS15, section 4
(M. Guyon, Inrap)
6After dismantling it became apparent that two planks of the flat bottom had been repaired (Guyon forthcoming). They cracked during installation, and so the carpenters chose to repair and consolidate them by transverse nailing (fig. 1) on the plank BS15, section 4. Another repair involved the replacement of a frame timber. This is visible because of the presence of a lot of small pegs, although this was very difficult to perceive before the wood was freeze-dried. We observed oblique alignments of three pegs (fig. 2), which exactly reproduced the alignment of nailing on the upper surface of the flat bottom under VP7. Once the old frame had been removed, the carpenters filled the holes left by the nails with the pegs before reattaching a new frame. During the archaeological excavations, no pegs were seen on the surface of the flat bottom: it was only during dismantling that these pegs were noticed and recorded. A further search for pegs was carried out after freeze-drying because during this operation a slight shrinking of the wood takes place making the pegs more easily visible.
Fig. 2: Pegs on the upper face of the flat bottom under VP7
(M. Guyon, Inrap)
7Another repair can be seen involving a frame timber. A hole had appeared in plank BS15, section 4, of the flat bottom because of a knot being removed, hence causing a potential leak. This hole was filled with caulking textiles held in place by a small wooden plate (fig. 3) (Rieth, Guyon 2010).
Fig. 3: Wooden plate on plank BS15, section 4 of the flat bottom
(M. Guyon, Inrap)
8Another repair was made using a piece of wood embedded in a frame timber. This was first thought to be a waterproofing repair but as nothing was visible on the flat bottom after its removal, the floor timber must have had a fault at this point and the carpenters preferred to remove it during construction. Anyway, this rectangular piece is not from the same wood.
9A major repair took place at the front of the boat. Several lead strips, originally from ancient piping, are present (Rieth, Guyon 2010). After removal of these lead strips, we saw that the widest strake had been repaired with many softwood wedges. This repair might have been as a result of operational wear on the plank, which is about 70 cm wide and had split naturally in its centre, which includes the middle of the tree, a fragile area from the outset. Alternatively the repair may have been necessary after a frontal impact, requiring a serious intervention including the cutting of many frames.
10At least nine wedges were found in the large crack in the central strake. Other wedges were present but have since disappeared because this part of the boat experienced a number of problems, particularly during its removal from the excavation site when parts of the flat bottom cracked during extraction. Once the wedges had been inserted, in order to make the repair as waterproof as possible, cloth coated with pitch was placed over the wedges before laying the lead strips. We clearly distinguished the weave of the textile used impressed in negative on the underside of the lead plate.
11Another repair filled a gap due to the missing part of the bilge, with a knee having a protuberance forming a heel on its lower face. Thanks to the dismantling it was also possible to observe manufacturing marks from a nail header on the underside of the nail heads.
12Similar observations have also been made on various shipwrecks, in particular on the Chalon-sur-Saône wreck (France), the Bevaix barge (Swizerland) and the Arles-Rhône 3 (France). These marks are not exclusive to nails for shipbuilding, but are also found on Roman shoe nails, for example. They could be interpreted as motifs recording the production of a particular blacksmith or blacksmith’s workshop.
13The waterproofing of the boat was made with textiles impregnated with pitch. During the manufacture of the flat bottom planks, carpenters gave a particular shape to the profile of the edge. This profile, once the two planks were positioned against each other, formed an inverted Y, with a slight protrusion downwards to keep the caulking material in place without any other means to secure it. The initial profile of the edge was axe cut. When replacing the waterproofing, the protrusion disappeared because, in order to remove it, a gouge had to be used to pull the pitch cloth out. Use of this tool eliminated the protrusion on the edge, thus allowing us to record precisely where the waterproofing had been replaced.
14During the dismantling of the barge 26 m of caulking material was removed for further study. It mainly came fro the space between the bilge and the side. At first, it looked like a piece of wood, but after a more detailed examination the textiles could be seen. During our research we were faced with a new challenge as to how we might carefully unfold this material and how this new data could be exploited to answer historical questions.
15Indeed, Roman nautical textiles are quite rare artefacts and previous studies have focused on the textile characteristics (Médard 2009, 2010a, 2010b, 2013, 2014, 2018; Schoefer 2004). Our aim was to link these textiles with the barge, to understand how the caulker did his job, and which techniques were employed.
16Until now, caulking textiles have been chemically unfolded by removing the pitch, leaving the cloth in a clean and flat condition for archaeological study (Plantec 2010; Schoefer 2004). Attempts to follow the previous procedure led to the destruction of the samples because of the very high pitch content, which had completely impregnated the textile. This unsatisfying state of affairs encouraged us to elaborate a new protocol, which began first with the question of the link between the material and the barge, and secondly how to record these operations, since each stage, from beginning to end, is a definitive transition with no possible return. The functional specifications of this new protocol were: to be efficient and fast; to limit waste, costs and toxicity for the operator; and still respect the textile. Organic solvents attack fibres, which become more fragile when treated this way and then tear at the folds. The new protocol used hot water to make the pitch plastic enough to allow the textile to be unfolded. With this technique, larger pieces of textile can be recovered (Meunier, Guyon forthcoming). Observation of impregnated textiles is more difficult because the pitch is not removed with the new method, but this technique allows for complementary recording.
17During the unfolding of the textiles, a clue appeared to identify the waterproofing technique used on the Lyon Saint-Georges 4 shipwreck. Normally on the Rhône-Saône axis between the 1st to the third century A.D. luting was used. This technique applies the waterproofing material in the course of boat construction, between the seams of timbers as they are being assembled. A very good example of this technique using pitched textiles, as on the Lyon Saint-Georges 4, has been recorded on the Arles-Rhône 3 shipwreck (Marlier 2014, p. 114). Marlier has shown that in this case large amounts of waterproofing material extend inside the bottom and that it can reach 6.5 cm outwith the seal between the bilge and the side. Excess waterproofing material can regularly be observed inside the boat. A lining was put over this to avoid the necessity of dealing with it. In the Arles-Rhône 3, the form of the seal was an inverted V with no protrusion. This can be explained by the use of the luting technique. Pitched textile was nailed during the construction, helping to keep it in place. This shipwreck, through her characteristics, date (first century A.D.) and location, can be compared to Lyon Saint-Georges 4. Here, no excess waterproofing material has been observed inside the shipwreck. The profile of the seals between the planks, the inverted Y, stops before reaching the other, interior side. This means that waterproofing material simply cannot physically escape beyond the seal. Moreover, while unfolding the pitched textiles, no lack of textile has been observed in the pieces located in the upper part of the seal, meaning that the excess textile extending beyond the seam might have been cut. And the protrusion at the bottom of the edge helps to keep the waterproofing material in place. All this led to the hypothesis of caulking instead of luting on this barge. This hypothesis was confirmed when we found an example from the area between the bilge and the side, in which a nail was trapped in the pitched textile (a second nail has also been found, showing this was not a unique example): the nail did not pierce the cloth (fig. 4). The textile is wrapped around the nail, and the fact that the textile is not pinned by the nail signifies that the waterproofing material was placed after nailing. The textile does not run perfectly to the end of the gap, meaning that there was no way of checking that it did. This shows that the waterproofing was done after the complete construction of the boat, and this process is caulking.
Fig. 4: Waterproofing material surrounds a nail
(M. Guyon, Inrap)
18Through exhaustive unfolding work, new perspectives and new data appeared. Nearly 3 m2 of Roman textiles have been recovered from the Lyon Saint-Georges, the most from one boat in Europe. Some vegetal textiles are part of this material, but they represent a small part of the discoveries, which are mainly animal fibres (essentially sheep and goat: it is very difficult to differentiate them). As regards the vegetal fibres, only two types of weaving are present: tabby and twill. This latter has been identified as a technical fabric already known in the third century as a support for sewing metal elements to form Roman armour (personal communication with J.P. Wild, April 2018, about his finds on the Vindolanda site). These textiles are of course reused, and we tried to find some clue as to their previous utilisation and to understand if they were specifically chosen for this task.
19Nautical textiles from ancient shipwreck have been found before now. The textiles coating the hulls of the Madrague de Giens shipwreck from the first century B.C. (Tchernia, Pomey, Hesnard 1978, p. 85) and the Nemi ships from the first century A.D. (Ucelli 1950, p. 265-268) were both woollen, and the thickness of the fibres used on the Nemi ships suggested a very high quality. However, no ancient source records this process. The fabrics we have found were tabby and were used between the planks of the hull and an external protective layer of lead sheeting. This indicates a specific production for that purpose, and a highly controlled quality. The use of wool in this location is a calculated choice, and after 2000 years it has come down to us, conserved by the resin that impregnated it. But these textiles were not the only ones used on-board. Sails and tarpaulin were also present, even if remains from ancient times are scarce. Black (Black 1996) assumed that since no piece of sail has been found in shipwreck excavations in the Mediterranean area, and since antique sources only indicate vegetal fibres as being used for that purpose, the material has not survived the ravages of time either underground or underwater. But with the tabbies of the Nemi and the Madrague de Giens, we have proof that nautical textiles have survived the ages, at least those made from wool when impregnated with resin, which are the same conditions as the pitched textiles used to waterproof the Lyon Saint-Georges’ shipwrecks, as well as those of Arles-Rhône and Tolozan.
Fig. 5: Warp-faced tabby from Lyon Saint-Georges 4, comparable to the Nemi’s finest textiles
(L. Meunier, IFIRES)
20Other nautical textiles have been found, mainly in Egypt, where the climate is more favourable to their conservation. These includes examples from Berenike, 1st to fifth century A.D. (Wild & Wild 2001), Quseir al-Qadim from 1st to second century A.D. (Handley 2011), Myos Hormos 1st century or early 2ndcentury AD (Whitewright 2007) and Thebes, from an Egyptian mummy conserved in Lyon, France, dated to the 2nd century AD (Schoefer, Cotta, Beentjes 1987; Rougé 1987; Wild & Wild 2001). All these textiles are from vegetal fibres. Those from Berenike were compared with those from the mummy by J.P. Wild, who identified several pieces as reinforcement strips, often seen as dividing lines on representations of traditional sails, forming a grid pattern where the horizontal widths of fabric intersect with brail lines, to which brail rings were sewn. Parts of these brail rings were found in Myos Hormos, Thebes and Quseir, thus helping to identify the strips on which they were sewn. In Quseir, the amount of material (2,455 fragments of Roman textiles) and the logic of utilisation by spatial location led Handley to assume that textiles remained in the place they were first used until eventual deposition, therefore, in a boat repair yard the cloth would come from a torn sail (Handley 2011, p. 323). This led us to reassess both new and old finds from the Lyon Saint-Georges shipwrecks. While making a catalogue of the nearly 150 pieces of textiles (Meunier 2018), it quickly appeared that some analogies could be made with the ancient nautical textiles from Nemi. The finest textile recovered was the fourth sample, a weft-faced tabby, with a count of 21/2x35 (warp/weft) threads in each direction and made of fibres with a 16µ range. This is a technical fabric that has been specifically chosen for the purpose by the Romans with no random element. Fifteen similar examples can be found in the collection of Lyon Saint-Georges, with an average fibre diameter of maximum 20µ and the same tabby aspect and z/z spun (fig. 5). Figure 5 is an example of this warp-faced tabby, with regular bundles of threads clearly designed to attach something to the fabric. The number of threads is only available for a few textiles because the survey is still ongoing, but it is between 24-28/10-12 and 90-110/16. Thread diameters are between 0.1 to 0.3 mm. Six textiles with these characteristics are also present in the collection of Arles-Rhône 3 barge (Médard 2013). In Lyon, three strips can be compared to those from Quseir, Berenike and Thebes, with a width of 4 cm, 3.77 cm and 4.5 cm. Even if the materials used are different – wool for the French textiles and vegetal fibres for the Egyptian – the textiles look the same, are from the same nautical context, and were adapted from material available in their geographical area. Previous surveys by Médard on half of the Lyon Saint-Georges textiles (Médard 2009, 2010a, 2010b, 2014, 2018) provided precious indications as to the technical aspect of the textiles and their fabrication, but made no real link with the nautical world. Médard’s experience in the study of clothing and domestic fabrics has convinced her that there are no sewn remains from that particular domain. Another detail to be noticed is the impregnation of the textiles with ochre, which cannot be dissolved in water or organic solvents, in 45 examples (also visible in fig. 5). This cannot be due to the burial phase, since the soil consisted of sand, or from the storage lake, because then all the textiles would have been impregnated. This can be interpreted in the light of nautical practices as a well-known technique recorded since antiquity: sails used to be treated to improve their technical performance and to avoid rotting. This technique has been successfully applied to the woollen sails of the Skuldelev ship replicas (Cooke, Christiansen, Hammarlund 2002) and also the sails of the Gyptis, a sailing replica of a 6th-century BC Greek sewn boat (Pomey, Poveda 2017), after two years because of the rotting of the sail provoked by dry/wet cycles (P. Poveda, personal communication). The textile is boiled in a mixture of seawater, grease and ochre, then rinsed in the sea and naturally dried. Similar aspects are present in the Arles-Rhône 3 collection. All these observations lead us to the conclusion that nautical textiles and certainly sailcloth were reused to waterproof the boats of Lyon and Arles (fig. 6). All of this survey is still ongoing and a doctoral thesis will begin next September focusing on the subject. A scientific publication on the new discoveries made during the conservation of Lyon Saint-Georges 4 is also in preparation.
Fig. 6: Example of the material that helped connect these finds to a nautical context
Two elements sewn together at right angles; two deliberately made circular holes of the same diameter; and a very thick (2.2 mm diam.) vegetal-fibre S2z sewing thread.
21While working on the newly found textiles, we had the opportunity to go and see the hemp sails of the Jeanne-Elisabeth, wrecked in 1755 near Palavas-les-Flots, France (Bartoš, Sanders 2012). Comparison with some of the nautical textiles conserved in Lyon showed the same reinforcement strip, with an approximate width of 5 cm, and with the same sewing points as on the LSG8 shipwreck, dated to AD 55. They are different on each side of the strip: one with wide stitches, spaced and with a triple twisted thread, and invisible sewing on the other side (fig. 4). This would seem to imply that we have a technological continuity in sail manufacture between AD 55 and 1755, that is, over 17 centuries. Given that nautical techniques are known to be conservative, this could be possible.
22Several complementary techniques have been used to record the different steps of the protocol. First, photogrammetry coverage of the caulking material was made in order to retain a 3D memory of how it was before and to provide documentation for recording. Thereafter, a GoPro film was made while unfolding the textiles, and a map of the different pieces of textiles was drawn. This was a transitional stage, where observation could be conducted.
23This stage was also recorded, and a relief map was made to understand how the textile was folded to be inserted between the planks. The folds are still visible and can be read by transferring the relief map into a monochrome picture. Through this technique, we compared the ways of folding the textiles. We wanted to know if it was always done in the same way or not, so that we might understand the actions involved in doing this task. We can also compare the textile type and the location of the samples to know if there is a link between the space to be filled and the textile used. This seemed to be correct, since the textiles used between the side and the flat bottom were often thicker and their preparation looked rougher, in that, the pieces used were less regular and transformed for this aim, than between the flat bottom planks, where the border was torn. These small pieces showed reduced dimensions to fit the space to be filled, in contrast with the space between the flat bottom and the side, which is wider. It seems logical to quickly fill this gap with a thicker textile, which would allow for faster work and which appears more suitable for this location.
24All these textiles can provide exceptional information on Roman techniques of caulking, of course, but also about textiles themselves. The way the caulker operated is still not well understood, but several avenues are currently being explored. We already have more than 2 m2 of unfolded textile, from nearly 6 linear metres of material, and there are still 23m waiting to be done. There is a great potential here to make a reference set of Roman textiles. This is what is planned for the near future.
25This work has shown the complementarity of, and the necessity of collaboration between the archaeologist and the conservator. The conservator’s know-how can shed new light on material and add to the understanding of artefacts. Repairs are small but can provide a lot of information regarding the maintenance applied to the shipwreck during its active life. Through archaeodendrochronology of these remains, we hope to know how long this boat was in use. For textiles, the conservation process has allowed us to recover information about textiles used to waterproof barges on the Rhône, showing that they originated within a closed system proper to the shipyard itself, even if complementary observations still need to be conducted. Textiles are more intimately linked to the artefact in which they feature, and are not just add-ons.
26Thanks are due to the Lugdunum Museum (Lyon, France) for allowing the study of the textiles from the Lyon Saint-Georges shipwrecks, to Inrap (Institut National de Recherches Archéologiques Préventives) for archaeological support, to ARC Nucléart (Grenoble, France) for technical support and access to the textiles, and to LPA (Lyon Parc Auto) for funding.