Mesolithic Freshwater Fishing: A Zooarchaeological Case Study

2.1 Freshwater Fishing in Mesolithic France

A bibliographical search enabled us to identify 21 Mesolithic sites in France in which bones of freshwater ichthyofauna were identified. Six additional sites only mention the presence of unspecified fish remains, that were most probably of freshwater or migratory species given their geographic location. This corpus thus comprises a total of 27 sites, distributed mainly inland over a large part of the country, and 32 assemblages for the entire Mesolithic period (Figure 1). More than half of these (15 out of 27) contain less than 25 remains and only 3 assemblages comprise more than 1,000 remains: the Cabônes shelter at Ranchot (Jura), the open-air sites of Noyen-sur-Seine “Le Haut-des-Nachères” (Seine-et-Marne), and La Chaussée-Tirancourt “le Petit Marais” (Somme), and probably also the Montclus site (Gard) (Table S1). This disparity is striking, especially as fine mesh sieving was carried out on most of these sites, albeit not always exhaustively (Table S1). In the valleys of northern France, where Mesolithic groups regularly settled on dry land near rivers, the presence of fish remains is not systematic even though the mammalian fauna is well preserved and exhaustive sieving was carried out (e.g. Ducrocq et al., 2019). Fishing is only clearly attested at a small number of sites. This is the case at Noyen-sur-Seine where, in addition to eel and pike remains, a dugout log-boat and fish traps have been preserved (Deseine, Guéret, Vigne, Mordant, & Valentin, 2019; Marinval-Vigne et al., 1989; Mordant, Valentin, & Vigne, 2013), and stable isotope analysis demonstrates that freshwater aquatic resources made a significant contribution to the human diet (Drucker et al., 2016).

Figure 1

Location of French Mesolithic sites having yielded freshwater fish remains (map of France from cartesFrance.fr). (1) Le Petit Marais, (2) La Haute-Île, (3) Le Haut-des-Nachères, (4) Baume de Montandon, (5) Bavans, (6) Châtaillon, (7) Les Cabônes, (8) Gigot, (9) Rochedane, (10) La Roche-aux-Pêcheurs, (11) Sous Balme, (12) Jean-Pierre 1, (13) Abri des Rocs, (14) La Doue, (15) Le Peyrat, (16) Bourouilla, (17) La Poujade, (18) Roquemissou, (19) La Baume Fontbrégoua, (20) Le Plaisir, (21) La Baume de Montclus, (22) Grotte du Salpêtre, (23) Le Mourre de Sève, (24) Balma de l’Abeurador, (25) Moulin, (26) Buholoup, and (27) La Crouzade. The Doubs River is highlighted in blue. Please refer to Table S1 in the supporting information for the data.

These findings motivated a detailed study centred on the ichthyofaunal assemblage of layer 3 of the Cabônes shelter at Ranchot in the Jura (Frontin, 2017), a region rich in Mesolithic sites with well-preserved faunal remains.

2.2 Regional Context

Settlement intensified during the Early Mesolithic, from 8400 BC onwards. Then, during the Middle Mesolithic, between 7600 and 7000 BC, an increase in the number of sites is recorded at all altitudes, including large open-air sites (Cupillard et al., 2015); occupation levels are denser, richer, and attest to diversified activities. At the beginning of the Late Mesolithic, the number of sites in the region decreases, possibly due to a climatic deterioration (i.e. the 8200 cal. BP event), after which the number of occupations increases again from 6000 BC onwards (Cupillard et al., 2015).

Mesolithic sites with ichthyofaunal remains are distributed along the entire course of the Doubs. This river flows across the Central Jura for 453 km, crossing the Bourgogne Franche-Comté region in France as well as the cantons of Neuchâtel and Jura in Switzerland. It is supplied by some 15 tributaries, and is the main tributary of the Saône, which is itself the main tributary of the Rhône. The flow of the Doubs is irregular, of the pluvial to pluvial-nival type with severe low water in the summer; the flood period extends from September to the end of May due to feeding by the heavy rainfall that affects the Jura massif. While the extent of the Weichselian alluvial deposits is well established (Fleury, Farjanel, & Collin, 1982), our knowledge of the postglacial evolution of the river is incomplete (Rotillon, 2000; Séara, Rotillon, & Cupillard, 2002). It is known that the rapid warming phases of the Lateglacial–Holocene transition impacted hydrosystems all over Europe, increasing river bed incision and changing systems from laterally active braided flows to meandering single channels (Macklin & Lewin, 2003; Magny & Bravard, 2002; Pastre et al., 2000). According to a study conducted at Choisey near Dole (Rotillon, 2000), the Doubs first incised the valley at the beginning of the Lateglacial, then river dynamics slackened in the Preboreal and Boreal periods. The development of riparian vegetation may also have played a role in fixing river courses.

The impact of climate change can also be seen on fish populations. During the Last Glacial, only cryophilic and low-temperature breeding species, such as salmonids, burbot (Lota lota), or pike (Esox lucius) and, to a lesser extent, rheophilic cyprinids such as common dace (Leuciscus leuciscus), were able to survive north of refuge areas (Mediterranean fringe and Black Sea basin) (Le Gall, 2010). Opportunities for recolonisation arose during the deglaciation process, and fish populations attained their maximum expansion during the Lateglacial period, expanding along the natural channels formed by the Danube and the Rhone. Cryophilic species then dispersed rapidly, followed by thermophilic species (Le Gall, 2010; Persat & Keith, 1997, 2002; Persat, 2003).

2.3 The Cabônes Site at Ranchot

Les Cabônes rock shelter (or “abri du Colonel Martin”) is situated in the last ranges of the Jura front. It is located on the right bank of the Doubs River at an altitude of 216 m, near the town of Ranchot (Jura), halfway between Besançon and Dole, in the lower Doubs valley (Figure 2). In this sector, the bed of the Doubs is wide and gently sloped (0.45%). The site is a small cave dug into a Lower Sequanian limestone ledge. It was probably formed by the joint action of the Doubs and a former karstic flow (Campy, David, & Cupillard, 1989; Cupillard & David, 1995). The opening of the cave forms a shelter about ten metres long. Following surveys carried out in the 1950s and 1960s, the site was extensively excavated from 1978 to 1989 under the direction of A. Thévenin, M. Campy, S. David, and C. Cupillard (Campy et al., 1989; Cupillard, 1998b; Cupillard & David, 1995).

Figure 2

General view of Les Cabônes site (Ranchot, Jura, France). The site is on the left, and the Doubs River is on the right. © C. Cupillard.

The stratigraphy is fairly well preserved in the front part of the site, where three complexes can be identified (Campy et al., 1989) (Figure 3):

1. A lower sterile complex consisting of pebbles lying on the limestone bedrock, corresponding to the Würmian alluvial deposits of the Doubs (layer 7).

2. An intermediate complex consisting of Doubs overflow silts incorporating angular cryoclastic fragments and blocks from the dismantlement of the porch (layers 4, 5, and 6). The top part of this complex (layer 4a) marks the end of the lateglacial allochtonous contributions of the Doubs. These layers are contemporaneous with Upper-Final Magdalenian occupations (David, 1993). The fauna consists mainly of reindeer, with a date for one bone of 13965 ± 101 BP, while a cave lion remain was dated to 12565 ± 50 BP, i.e. 12620–13201 cal. BC (Cupillard et al., 2015; Stuart & Lister, 2007).

3. An upper complex composed of clayey matrix with cryoclastic fragments, which is a local expression of the “groizes litées” resulting from the dismantling of the shelter. This complex is subdivided into two layers:

Figure 3

(a) Stratigraphy of Les Cabônes rock shelter. After S. Roué (2000). © N. Carquigny and (b) view of the excavation to the east with annotated stratigraphy. © C. Cupillard.

Layer 3 (formerly AOC, for Cryoclastic Organic Clay) is black and highly organic, rich in micro-charcoals of anthropic origin. It is 60–70 cm thick, thinning towards the Doubs around squares 6–8. It was excavated over 50 m² for a potential surface area of 80 m². The disturbances (burrows and ancient excavations) are generally located at the bottom of the shelter up to the porch. The outer zone of the site where material is very abundant probably corresponds to one or more discarding areas on the periphery of a dwelling area. However, little evidence of the latter subsists, apart from a hearth marked by a rubified zone in squares Z-A 14 (Figure 4). A lithic industry attributed to the Middle Sauveterrian forms most of the corpus; Early Mesolithic microliths such as crescents and oblique blunted points are also present at the base of the layer (Crotti & Cupillard, 2013; Cupillard et al., 2015; Roué, 2000). Isolated human remains were collected in this level, notably a skull fragment from square D9 dated to 8985 ± 40 BP (GrA-38019) belonging to a female individual related to haplogroup U5b1 (Posth et al., 2016) and the Villabruna cluster (Fu et al., 2016). Evidence from wild boar, red deer, and roe deer age-at-death data indicate a multi-seasonal use of the shelter (Leduc, Bridault, & Cupillard, 2015; Trémolières, 2020). The site might have been used (and reused) as a base camp as indicated by the abundance and diversity of the material recovered (Crotti & Cupillard, 2013; Cupillard, 2003). Layer 3 probably corresponds to an accumulation of several occupations dated between 8290 and 7060 cal. BC (2 sigma) (Table S2).

Figure 4

Excavation map of Les Cabônes shelter at Ranchot and sampling location of analysed fish remains from Layer 3; scale: a square represents 1 m². CAD A. Bridault, modified from Leduc and Carquigny (Leduc et al., 2015).

Layer 2 (formerly AC, for Cryoclastic Clay) is light-brown in colour and thickens towards the Doubs. It is largely disturbed by burrows and contains rare Mesolithic flints as well as Neolithic, Protohistoric, and Gallo-Roman remains. Radiocarbon analyses of two red deer bones yielded dates between 7570 and 7180 cal BC (2 sigma) (Cupillard et al., 2015; Drucker, Bridault, Hobson, Szuma, & Bocherens, 2008).

A sterile grey clay layer (Layer 1) and a surface layer (Layer 0) seal this complex.

It is noteworthy that the sedimentary nature of the upper complex (layers 0–3) points to a very weak or non-existent influence of the river in the formation of the Mesolithic archaeological levels. This contrasts with the lower silty levels (4, 5, and 6) which, however, did not yield any fish bones.

3.1 Materials

The ichthyofaunal remains were collected in layer 3, after fine-mesh sieving (nested sieves of 10–1 mm) of all the sediments extracted during the excavation. The studied sample comprises a total of 9,328 remains grouped into 2 differently sorted subsets:

1. A first sample consists of 3,361 remains collected over a surface of 40 m² and sorted with the naked eye and/or under a magnifying glass (2–5×) by the archaeologists in the 1980s and 1990s. A preliminary study was carried out on 438 of these by P. Morel who identified the presence of at least five species (Morel in Cupillard, Auguste, Cupillard-Perrenoud, David, & Morel, 1988): brown trout (Salmo trutta), chub (Squalius cephalus), barbel (Barbus barbus), perch (Perca fluviatilis), grayling (Thymallus thymallus) and, possibly, burbot (Lota lota) and pike (Esox lucius). This sample also contains remains collected on sight, mostly large Salmonidae remains.

2. The second sample was sorted by D. Frontin, a fish specialist, under a stereomicroscope (10–40×). It comprised 5,967 remains from five squares (H8/9, I9/10, and J10), equivalent to a volume of 13.63 L of sieving refuse resulting from the combined fractions of the sieving column (effective screen size 1 mm) (Figure 4). This area was selected because of the good preservation of the remains (Cupillard & David, 1995).

A small quantity of recent remains discarded on the site by a pair of kingfishers (Alcedo atthis) nesting in one of the shelter walls were mixed with these archaeological remains, but could be excluded from the sample because of their “fresh” appearance. Because of this, particular attention has been focused on determining the origin – anthropic or otherwise – of the archaeological ichthyofaunal remains.

3.2 Methodology

4.1 Taxonomic Composition of the Studied Assemblage

The comparison of the faunal spectra resulting from the two sorting methods (by archaeologists and by the specialist) showed no difference in the representation of the identified species (Frontin, 2017). The results are therefore presented altogether.

Six families and seven species make up the Ranchot Layer 3 spectrum (Table 1). Cyprinid remains are largely predominant (96.6%), of which three species are attested with certainty – common bream, roach, and minnow.

The latter were identified using several diagnostic elements (basioccipital, second vertebrae, and cranial elements), only slightly affected by fragmentation.

The other families are represented by only one species each - trout with 243 remains (i.e. 2.6% NISP), grayling, burbot, perch, and eel, with only a few remains each (i.e. less than 1% NISP).

The species making up this group have fairly similar ecological habits. They are mostly gregarious and diurnal, usually live close to the bottom or in open water, and circulate in rather cool, well-oxygenated, and relatively running waters. The eel, the only migratory species, also adapts to these diverse environments. The predominance of cyprinids in the assemblage undoubtedly reflects an environmental reality. It is the most represented family in the environment in terms of number of species (Keith & Allardi, 2001).

4.2 Fragmentation Rate of the Material

The state of fragmentation of the material is estimated using a percentage of completeness assigned to each specimen (Butler & Schroeder, 1998; Nicholson, 2000). For fish vertebrae, which are the most abundant bones in the assemblage, the condition of the centrum is recorded, as spines are often absent in archaeological material. Less than 10% of a sample of 1,184 vertebrae are less than 50% complete (Table 2). Consequently, fragmentation is low and has little effect on identification. On the other hand, the opposite configuration was observed for the pharyngeal arch, an element of a masticatory apparatus in cyprinids, which is quite resistant but often found fragmented. The presence or absence of teeth has not been taken into account, as they are easily detached. Only complete or well-preserved arches, which represent a minority (11 are 50% complete or more), were used in order to produce a reliable diagnosis.

4.3 Bone Modifications

Overall, bone surfaces are well preserved and only 13% of the remains show modifications (Table 3). No pitting marks, characteristic of the attack of bone surfaces by digestive juices, were recorded in the assemblage. Bone rounding caused by sediment abrasion is the most common modification found on any bone element, for all species and all sizes of specimens. It is clearly differentiated from rounding due to transit, on account of the absence of the characteristic rounded and glossy appearance of the latter (Llona & Andrews, 1999), which is more often confined to the extremities of elements and does not extend over entire surfaces. Only two types of deformations were found on vertebrae: “compression” and “undulations.” Compression of the articular faces, or even more rarely of the entire vertebral body, was only observed on small specimens (N = 46), mainly cyprinids and on a few salmonids. Two vertebrae presented crushing. The first shows crushing associated with a torsion of the vertebral body and the other displays crushing associated with a deformation of the lateral decoration. All of these deformations are related to traces of mastication described in the literature (Butler & Schroeder, 1998). These bones were chewed, but not ingested; instead, they were spat out after the flesh was consumed. This would therefore indicate that small fish were of dietary interest. The second type of deformation takes the form of “ripples” on the edges of articular faces (N = 44). This was only observed on white-coloured bones, which seem to result from heating to very high temperatures (calcination). Finally, changes in bone colour (from black to white) indicate different stages of heating, from carbonisation to calcination. These burn marks are observed on about 4% of the remains of all types of bones, all species, and all sizes of individuals taken together. They are probably not related to the cooking of fish but more logically to the burning of bone waste.

4.4 Representation of Skeletal Parts

The distribution of bones of all taxa into three anatomical units shows that axial skeleton bones are predominant (77.2%), mainly vertebrae. The presence of elements attached to the vertebral column (ribs, Weberian ossicles) is very marginal (Table 4). The appendicular skeleton is only documented by a few fin rays (0.1%). The cranial elements (basioccipital, pharyngeal arches, and teeth) account for 22.7%.

Only cyprinids are represented by different skeletal parts, including head bones. The perch is only represented by a few cranial remains, while the other species are only represented by their axial skeleton. The vertebrae are significantly more abundant at Les Cabônes (76.4% of NISP) than expected in a complete cyprinid skeleton (23%);^[4] head and appendicular skeleton are significantly under-represented (χ ² = 987.81, p < 0.0001). How can this distribution of anatomical units be interpreted? As no traces of incision were observed on the bones, it is impossible to assess whether the fish were cut up before consumption. The under-representation of the cranial elements could reflect heading before being brought to the site, but in view of the small size of the catches (see below), this seems unlikely. On the other hand, their destruction in situ by trampling and/or weathering is plausible. Most of the cranial elements have a lower bone density than post-cranial elements (vertebrae in particular), and are not very resistant to destructive factors (Butler, 1996; Butler & Chatters, 1994). Teeth, the basioccipital, or pharyngeal arch are the most resistant cranial elements, and often survive best in archaeological material (Le Gall, 1982, 1984; Wheeler & Jones, 1989). The differences observed in the degree of fragmentation of pharyngeal arches and (more complete) vertebrae are consistent with this (Table 3). Moreover, the remains of salmonids and thymallids are more degraded here than the remains of other taxa, which is in keeping with taphonomic logic.

At Ranchot, the most resistant elements are thus the most abundant: vertebrae for all species, basioccipitals and pharyngeal arches for cyprinids, and teeth for cyprinids and percids. Consequently, it is likely that the differential representation of the skeletal parts of this assemblage results mainly from the in situ post-depositional destruction of preparation and/or consumption waste.

4.5 Characterisation of Captured Individuals

The relative representation of species in terms of the number of individuals caught was not assessed in MNI, due to the characteristics of the assemblage, but through %NVcor (refer methodology section) (Table 5). The method confirms the strong predominance of cyprinids (97%) established on the basis of the count in number of remains. In comparison, the other species, even the best represented species such as trout, represent an anecdotal number of catches.

This very large number of cyprinids is somewhat tempered by the small catch size. The reconstruction of the total length (TL) of individuals on a sample of 185 cyprinid vertebrae shows a distribution of values between 9.9 and 21.1 cm, with a median of 14.5 cm. Small cyprinids therefore appear to have been targeted. According to Fishbase (Luna, 2021b), the common size of a roach is 25 cm, and the average size of common cyprinid species potentially present in the assemblage is around 20 cm. The individuals from Ranchot therefore appear significantly smaller than expected. The size distribution of catches (Figure 5) forms a quasi-symmetrical bell curve, suggesting random recruitment of individuals.

Figure 5

Reconstructed size distribution of cyprinids from layer 3 (N = 185 vertebrae).

This pattern also seems to be observed among certain rare species in the assemblage. Grayling shows a median size of 17.5 cm at Ranchot, for 17 vertebrae, whereas individuals caught today are commonly 30 cm in length (Luna, 2021c). Due to the small sample size, the distribution is not very precise, but shows a predominance of individuals between 10 and 25 cm, which is consistent with data for cyprinids (Figure 6). Similarly, the four reconstructed remains of burbot show sizes between 17.4 and 25.4 cm, for a median of 20.1 cm, well below the commonly expected size of 40 cm (Luna, 2021a). Finally, the only reconstructed perch remain is also from a small individual, estimated at 14.5 cm for a common species size of 25 cm.

Figure 6

Reconstructed size distribution of grayling from layer 3 (N = 17 vertebrae).

Salmonids show a different pattern. On the one hand, reconstructed individuals (165 vertebrae) are larger and much closer to expected sizes, with TLs ranging from 6.6 to 64.2 cm, for a median of 33.5 cm. Today, the common size of a brown trout is 35 cm, even if record-breaking individuals can reach 1 m (Rochefort & Corolla, 2019). The reconstructed individuals are therefore consistent with these observations. On the other hand, size distribution here is anything but symmetrical (Figure 7). The curve gradually increases from 5 to 40 cm, reaches a mode in the [35–40 cm] class, which alone accounts for 22% of the catches, and then falls sharply beyond that, with only 6% of the catches measuring between 40 and 45 cm. There thus seems to be a threshold effect limiting the presence of large catches.

Figure 7

Reconstructed size distribution of salmonids from layer 3 (N = 165 vertebrae).

The eel could show a similar pattern. The sizes of the only two reconstructed individuals are 40.5 and 41.5 cm, in line with the common size of the species, 40 cm (Froese, 2021).

Based on data obtained for each species, we can also propose a reconstruction of catch sizes for the collection as a whole. To do this, we divided the NVcor of each species into size classes proportional to those found in the samples used for size estimates, and then added them together. The resulting projection, shown below in Figure 8, clearly illustrates the predominance of small cyprinids in the assemblage: almost 94.5% of the catches are less than 20 cm in total length, 99.3% of which are cyprinids. Medium to large fish are thus very poorly represented, with only 2.1% of catches exceeding 25 cm, 95.4% of which are salmonids.

Figure 8

Projected size distribution of reconstructed fish sizes, all taxa, from layer 3, N = 5,158 (size distribution patterns observed for each species have been applied to total corrected number of vertebrae from the species).

4.6 Food Contribution of Species

Using the measurements taken to reproduce catch size, it is also possible, on the same sample, to estimate weight. We were thus in a position to calculate an average specimen weight for each species in the assemblage, varying between 35.4 g for cyprinids and 495.7 g for trout, as presented in Table 6. In order to assess the relative contribution of species to the diet of the Ranchot occupants, these values were then multiplied by the NVcor of each species. The results of this transformation being strictly proportionate to the number of captured individuals, they allow the unbiased estimation of the part of each species in the fish eaten on site. The exact figures for the weight provided by each species are nonetheless theoretical constructs, and must not be taken at face value.

The results show that although cyprinids still dominate, trout made a significant contribution to the diet, accounting for more than a quarter (28%) of the consumed fish, despite the small number of remains (2.71% of NVcor).

4.7 Fishing Seasonality

Due to the very small size and state of preservation of remains, only 82 vertebrae had the required characteristics for a skeletal-chronological analysis of the season of capture. Of these, only 57 (70%) provided the three concordant readings required to ensure analysis reliability (sample details in Supporting Information, Table S4). Our study of catch seasonality is therefore based on a relatively limited sample. The results nevertheless show a strong structuring of the data (Figure 9): more than 57% of the catches occurred at the beginning of the good season. The number of catches then decreased rapidly as the summer season progressed, and only 7% of individuals appear to have been caught in the bad season. Thus, although fish could be caught all year round in Ranchot, it seems that this resource was mainly exploited in early spring, a period of food shortage known to be difficult and poor in resources.

Figure 9

Distribution of identified season of death on a sample of 57 fish vertebrae from layer 3; GS = good season, BS = bad season.

No structuring of the data by species was observed.

5.1 Origin of the Bone Accumulation

The question of the origin of the bone accumulation is central to any ichthyofaunal study, but was particularly central at Ranchot. The observation of present-day kingfisher pellets during the excavation immediately raised the possibility of a natural accumulator, and the unusually small size of the remains did not help to dismiss this hypothesis.

Specific attention was therefore paid to this problem during our study, in particular, through the analysis of a sample of remains from kingfisher pellets, leading to a better characterisation of this type of accumulation. A summary of the main results of our observations is presented in Table 7 below.

Fortunately, these data exclude the intervention of this accumulating agent with almost complete certainty. If we consider the Ranchot assemblage as a whole, it is clear that all the larger individuals, with vertebrae larger than 2 mm in diameter and/or with a reconstructed size larger than 15 cm, could not have been the prey of such a small bird. However, a mixture of kingfisher regurgitation with another accumulation process remains possible, and we thus carried out a more detailed comparison between the characteristics of the cyprinids from Ranchot (which overwhelmingly dominate the smaller remains) and those of the fish consumed by the kingfisher. This comparison shows that although the cyprinid remains from Ranchot are small, their average diameter is significantly greater than those from the kingfisher pellets (Student’s t = 8.42; p ≪ 0.001), and it is statistically unlikely that they are of the same origin. Above all, the reconstruction of sizes based on this sample shows that the vast majority of cyprinids captured at Ranchot, with an average size of 14.9 cm, exceed the consumption capacity of the kingfisher. Out of a total of 185 fish for which size could be estimated, all but one^[5] exceed 10 cm, the maximum size of prey listed in the literature (Hallet, 1985; Libois & Hallet-Libois, 1995), and 61.4% exceed 13.5 cm, the maximum size observed in the reference assemblage. Finally, the unimodal, symmetrical bell-shaped size distribution curve of the Ranchot cyprinids (Figure 5) is more suggestive of a single origin of the population than a mixture of two populations resulting from distinct selection processes.

It is also possible to exclude most of the other potential accumulative agents of bone concentrations. Relatively few piscivorous predators can be found in caves and rock shelters, and mammals give rise to very different assemblages from those of Les Cabônes. The catch size observed at Ranchot is not incompatible with otter habits (typically from 10 to 30 g for cyprinids and percids (∼10–20 cm), and from 100 to 200 g for salmonids (∼30 cm) Guillaud et al., 2014), but the remains consumed by this species are significantly marked by digestive processes, with 90% of the remains bearing acid attack marks, resulting in polishing, pitting or surface dissolution, and tooth marks and deformations affecting 3% of the remains (Guillaud et al., 2014). As no digestion marks are observed at Les Cabônes, otter action can therefore be excluded. The dog produces assemblages that are even more marked by taphonomic degradation, with very intense fragmentation of remains, and often unrecognisable surviving bones, reduced to corroded fragments and marked by traces of digestion (Jones, 1986, 1984; Nicholson, 1993). Again, this appears to be incompatible with the state of preservation of the Ranchot remains.

Fish bone assemblages generated by the Eurasian eagle-owl (Bubo bubo) similarly display features inconsistent with that of the Ranchot assemblage. First, the bones ingested by owls are marked by digestive processes, though to a lesser extent than by mammals: about 9% of the remains display moderate to intense digestive alterations (Guillaud et al., 2018). Second, the eagle owl tends to select prey significantly larger than those found in our archaeological assemblage: Guillaud notes that the predominant weight class in the eagle owl sample included fish estimated between 200 and 300 g (Guillaud et al., 2018), whereas the average weight in the Cabônes sample was no higher than 48 g. Finally, no piscivorous specialisation of the eagle owl has to our knowledge been described (Russ, 2010), and assemblages generated by the species should therefore include many remains of birds and small mammals, mostly absent from the archaeological sample.

Finally, there remains the case of the barn owl (Tyto alba). This nocturnal raptor has an opportunistic diet that can in specific circumstances become primarily piscivorous. The bird shows a general preference for prey weighing 10 and 70 g (Broughton et al., 2006), consistent with most of the fish consumed at Ranchot. Pitting and deformations are observed on less than 5% of the remains, but 19% show traces of rounding, which has not been observed in the archaeological collection. Moreover, the piscivorous specialisation of the barn owl has only been recorded in North America, in the case of the seasonal abundance of dead or dying fish stranded in the pools of an intermittent stream (Broughton et al., 2006). These environmental conditions are far different from those found on the banks of the Doubs, a deep and wide permanent river, and the estimated season of capture (early spring) is similarly conflicting. Therefore, a broader-spectrum diet is to be expected for the barn owl, and, as for the eagle owl, the associated presence of small mammals should be expected were the accumulation produced by this species.

The inconsistencies between fish bone assemblages produced by animal accumulators and that of Les Cabônes appear therefore significant enough to exclude an animal origin of the archaeological collection. A natural origin can likewise be rejected: the stratigraphy of the shelter shows almost no influence of the Doubs River in the formation of the sediments of the Mesolithic levels having yielded fish remains. It seems therefore difficult to argue in favour of natural inputs of fish carried into the shelter during seasonal river flooding. Less importantly, the presence of the bream possibly upstream of its natural habitat also contradicts a natural fluvial origin. Finally, the size distribution pattern of salmonids, displaying a relatively brutal threshold for fish over 40 cm, appears inconsistent with a natural accumulation.

We therefore consider that an anthropogenic origin of the archaeological fish assemblage of Ranchot can be accepted, even though a marginal contribution by other accumulators cannot be ruled out entirely.

The spatial distribution analysis of fish bones shows a fairly homogeneous overall distribution of skeletal elements with, however, denser accumulation zones (Z-A-B 10–12, C10, H-I-J 8–10) coinciding with the preferential accumulation zones of lithic remains (Roué, 2000). Nevertheless, it is noteworthy that Salmonidae remains seem to be preferentially discarded near the hearth. Masticated and burnt remains are also preferentially located near the hearth.

5.2 Prey Acquisition

5.2.1 Fishing Environments

Fish are particularly sensitive to the physico-chemical and biological conditions of their living environment, and each species has specific environmental requirements. The slightest variation in these conditions influences its presence and demography in a given place. In a river environment, we observe a longitudinal organisation from the source to the mouth, in zones associating environmental conditions and well-defined fish populations. The most commonly used typology, proposed by Huet in 1949, defines five ecological zones with their associated species (Huet, 1949). Based on this range of species, it is thus possible to evaluate the capture zone(s) of individuals with an ichthyofaunal spectrum.

We thus plotted in Figure 10 the ecological preferences of the fish species identified at the site on the ecological zones of the Doubs (terminal zone not represented) (Verneaux, 1973).

Figure 10

Ecological requirements of the fish species identified at Ranchot, plotted along the longitudinal zonation defined by Huet (1949), from upstream (Trout zone) to downstream (Bream zone). Total: number of Ranchot species present in sub-zone; pink area: minimal river area including all of the species documented at Ranchot.

The first observation is that, according to these data, it is highly unlikely that all the fish in the shelter were caught in a single location. The requirements of minnow and brown trout are incompatible with those of the two breams in terms of current, oxygen, and temperature. Fishing therefore seems to have been organised into river sections, with catches in different places. At the very least, for the capture of all the identified species, the fished area had to cover the downstream half of the grayling area and the upstream two-thirds of the barbel area. In addition, incursions into more distant fishing grounds cannot be excluded. Nevertheless, species composition remains consistent: 11 of the 14 determined species are typically present at the transition between the grayling and barbel zones, which could indicate a preferential fishing area.

The species identified in the corpus, and in particular cyprinids, generally prefer shallow waters with moderate currents. These species are therefore mainly found on the banks of the main channel or in secondary arms. Only grayling and the perch show divergent ecological preferences, favouring, respectively, the benthic zone of the main channel and slow-moving backwaters. These two minority species in the corpus may however move to secondary arms during the reproduction period. This points to the existence of a preferential catching environment for the whole assemblage.