Gaelic Folklore (34): Sanctuaries and Sacred Places

Sanctuaries and Sacred Places 

Drawing on iron age texts and a critical examination of archaeological data, it is possible to suggest some key features of the sacred spaces among the Celts.

Written Evidence

As the iron age Celts made limited use of writing, textual accounts of Celtic peoples derive almost entirely from Graeco-Roman sources. Those classical texts contemporary with the Iron Age are most relevant to a discussion of iron age sites, and the present account concentrates on these. As the product of an external, conquering society, classical commentaries are in many ways a problematic source of data on the Celts. But at the same time, such accounts are iron age artefacts, and when treated as such can yield much information.  Most classical accounts of Celtic sacred space date to the later Iron Age (i.e. La Tene D; C.120-0 Bc). Greek colonization in the western Mediterranean, under way by the sixth century Bc, and Celtic expansion into northern Italy from the fourth to second centuries Bc and Asia Minor from the third generated limited textual data on the Celts, but it was only from C.125-120 Bc, with Roman intervention in the Provincia, that significant quantities of data emerged on Celtic peoples and practices. The date of these texts ensures that the question of Graeco-Roman influence, both on Celtic practices and on their literary depiction, is ever-present in assessing the literary evidence. 

Underlying most iron age texts is thus a common theme: a sacred area comprising or including an enclosure. How such enclosures were realized is nowhere specified, but the concept of sacred enclosure is clear. Ironically, given their interest to the excavators of the weapon-rich sites discussed below, almost the only texts for which enclosure remains an uncertainty are Caesar's references to loci consecrati (sacred spots). Caesar uses the term for the druidic meeting-place among the Carnutes of Gaul and for heaps of battle spoil dedicated to 'Mercury'. In the latter case, Caesar simply says that booty was piled up in heaps; enclosure need not be implied. Elsewhere, Diodorus referred to booty being taken off the battlefield, without specifying its fate. Livy suggested spoils were heaped on the battlefield, apparently ruling out formal structure, but wrote with reference to the Celtic sack of Rome in 390 Bc. As Brunaux suggested, battlefield booty heaps may have been used in foreign territory, when spoils could not be moved to existing loci. Whatever the case, it is clear from Caesar's account that in later iron age Gaul booty heaps were conceptually bounded by their taboo status. The almost total absence of any detail on construction and use make it difficult to determine the nature of Celtic cult sites. But the formalized structuring of space, incorporating enclosing works, is suggested by most texts.

A 'Natural' Religion?

'Natural' loci such as woods, groves and trees are often advanced as Celtic cult loci par excellence. This concept of Celtic sacred space derives in part from classical texts, but references on this theme do not commence before the first century AD, when woodland sites are frequently associated with the druids. Iron age references to 'natural' loci are rare. In C.100 Bc Artemidorus commented on an open-air ritual at Cape St Vincent in western Spain, and Poseidonios on the use of lakes as sacred repositories in the Toulouse area of the Provincia. Strabo also stated that the Celtiberians worshipped a nameless god outside their houses.

Although only Artemidorus makes explicit reference to the absence of structures, all three texts may imply rituals without use of formal (roofed) structures, or otherwise involving the restructuring of space; for example by construction of an enclosure. There are no other iron age references to such 'natural' cult sites.

The first-century AD change in emphasis to woodland sites is ignored by archaeological commentators except Wait, although we can advance reasons for a change in practice at this point. Imperial proscription of the druids could have forced the use of secluded locales. Equally, as Chadwick noted, the textual association of druids and groves could be the result of a spurious etymology, voiced in Pliny's suggestion that 'druid' derives from the Greek word for oak.  It is not impossible that the association of groves with Celtic religion is a literary construct of the first century AD.

The concept of nature-based Celtic religion is also predicated on the apparent paucity of archaeological evidence for tectonic loci. As such sites would leave little archaeological trace, this proposition remains difficult to test, but it is important to note that the archaeological record is likely to be biased in favour of formal structures. Bearing this in mind, what evidence can be adduced for 'natural' loci?

Nemeton

Nemeton is a Celtic descriptor for sites of cult significance. Nemeton is attested mainly in post-conquest place-names and epigraphy, over a wide area. Place-names include Augustonemetum (Clermont-Ferrand) and Nemetodorum (Nanterre) in Gaul, Nemetobriga (near Ortense) in Spain, and Medionemeton (near the Antonine Wall) in Britain. Epigraphic attestations include Mars Rigonemetis from Britain (Nettleham) and Nemetona from Altripp (Spier). Nemeton is commonly glossed 'sacred, with little indication that it principally denoted such sites. It is suggested nemeton had several meanings, including grove or clearing and small shrine or chapel (d. the Irish nemed, glossed sacellum, suggestive of a small shrine). Formal structural associations are a possibility for some examples, as at Vaison-la-Romaine (Vaucluse, France), where an inscription records the dedication of a nemeton to Belissama. and the context suggests a structure or precinct. Nemeton thus remains enigmatic. The most acceptable gloss, MacCana's 'a sacred place', simply emphasizes how little we may infer from the term regarding the form of Celtic cult sites.

Trees and Groves

Groves continue to be advanced as Celtic cult sites, essentially by reference to nemeton attestations, and the first-century AD texts discussed above. Such sites would have little archaeological visibility, but some recent discoveries may suggest that representations of trees were employed in cult contexts. These are the finds of bronze leaves at the St-Maur sanctuary (Oise, France) and the remains of a golden tree from a pit at Manching (Bavaria). These finds are problematic; the Manching tree, which probably dates to the third century BC, and from which were suspended golden leaves and fruit, is apparently of Greek manufacture. The cult association of the find, which is not from one of the possible Manching cult foci, is also uncertain. Iron leaves, again designed for suspension, were found in a settlement context at Villeneuve-St-Germain (Aisne, France). The positioning and use of such finds within sanctuaries (where related to these) are as yet unknown.

Water

'Watery' loci such as a springs, lakes and bogs have traditionally been argued as important sacred spaces for Celts.

Springs

Springs are commonly seen as a focus for Celtic rituals. Textual evidence in this respect is very sparse. Hirtius' account of the diversion of a spring at Uxellodunum in 51 BC, and the subsequent surrender of the oppidum is a possible reference of iron age date, but a utilitarian function could well be implied for this spring. The one reference to springs in a clear iron age cult context is Lucan's overwrought account of a Lucus near Marseilles, written in the first century AD with reference to the Civil War.

Archaeological evidence for springs as pre-conquest cult foci is virtually non-existent. The argument for Celtic use of such loci is heavily dependent on the well-documented post-conquest use of springs for cult purposes. Finds of wooden statuary (a non-classical medium) at the Cote d'Or sites of Sources de la Seine and Elssarois and at Chamalieres (Puy-de-Dome), and elsewhere in France, have been argued to show a Celtic presence at springs, but can nowhere be dated before the first century AD. The recently excavated examples from Montlay-en-Auxois (Cote d'Or) were found in wooden catchment pools dated by dendrochronology to AD 86-119.

On the other hand, cult activity occurred from the early post-conquest period at many springs: Luxeuil and St-Marcel (Indre), Avord and Sagonne (Cher), Vichy (Allier), Chateauneuf-Ies-Bains (Puy-de-Dome), and Coren and Vic-sur-Cere (Cantal) have all produced early post-conquest coins or ceramics. There are two ways to consider this phenomenon: either by the retrospective inference that such sites are close enough in date to the Roman intervention to suggest they represent the survival of pre-conquest practices; or as a veritable post-conquest phenomenon. On present evidence, the balance is in favour of the latter.

Lakes

Poseidonios, who visited Mediterranean Gaul in C.IOO BC, refered to sacred lakes as repositories for treasures among the Tectosages of Toulouse. This passage is often cited as evidence for a widespread use of sacred lakes by Celtic peoples, but the text is clearly specific to the Tectosages. Archaeological evidence for lakes in cult contexts is, however, more widespread. The deposition of high-value metal in watery contexts, including lakes, is well attested archaeologically in Atlantic Europe. This practice increased during the Iron Age in Britain, though on the Continent it declined after the second century BC. Numerous lakes are interpreted as depositional cult foci, including Llyn Cerrig Bach (Anglesey) and Carlingwark (Scotland). A similar interpretation was originally advanced for the lake site at La Tene, (Neuchatel, Switzerland), where thousands of weapons and tools, and some jewellery and coins, were found.

Bogs

Bogs also served as foci for metalwork deposits. This practice was not restricted to Celtic peoples, and features for example in Germanic cult. The Gundestrup Cauldron, widely seen as the quintessential 'Celtic' cult artefact, was in fact found in a bog in Himmerland, Denmark.  Human remains are mainly known from Germanic contexts, but sometimes occur in Britain and Ireland. The Lindow bog body (Lindow Moss, Cheshire) is a recent example. Dating of the body is problematic, but radiocarbon dates from the most recent analysis cluster around the first century AD. Lindow man suffered a threefold death (by axe blows, garrotting and cutting of the throat). This triplication suggests a death with ritual links. Where datable, however, British bog bodies are mainly of bronze age or Roman date, and their ritual associations unclear. The extent to which such deposits represent an iron age ritual phenomenon is thus uncertain. 

There is some evidence that islands were favoured as cult sites by Atlantic Celtic peoples. Several texts may be noted in this context. Posidonius described a Gallic site on an island off the mouth of the Loire, served by female religious specialists. Tacitus, describing Suetonius' attack on Mona (Anglesey), associated the island with the druids and with women whom he compared to the Furies. After the conquest, Mela referred to virgin priestesses on the island of Sena (Brittany). Female religious specialists are linked to islands in all three passages. The implication is possibly of sexual boundedness. Islands are also physically bounded, relating them conceptually to other forms of enclosure noted here. In Britain, archaeological evidence for an island sanctuary occurs in the form of a wooden circular temple on Hayling Island.

A link theme in the above discussion is that of water as a boundary. Brunaux has suggested that lakes were natural sanctuaries, without enclosure or protection. This may not always have been the case, but water, or islands in water, clearly offer a ready-made form of enclosure. Both textual and archaeological evidence for iron age 'natural' foci is thus very restricted, the latter being largely limited to certain water categories. It is very likely that Celtic peoples did employ natural forces as cult foci, without restructuring their space. In the Graeco-Roman world, for example, such loci co-existed with formal, public, cult centres and the same is clearly probable for Celtic peoples. Indeed, it is possible to argue that our concept of Celtic religious life is distorted by the poor recognition of 'natural' cult loci by both classical commentators and archaeologists. In this context, the lack of textual data may reflect the fact that non-structural loci would not have been easily recognizable to external observers.

But at the same time, natural forces, including trees and springs, were familiar cult loci in the Graeco-Roman world.  It remains significant that almost all iron age textual references suggest, with varying degrees of certainty, that the designation of Celtic cult sites involved the restructuring of space. Equally, while the difficulties of archaeological recognition of 'natural' sites must be borne in mind when assessing their importance to the Celts, it is also reasonable to expect some positive archaeological evidence in their favour. In many of the cases examined above, this is clearly lacking.

Wells and Shafts

Wells and shafts have long been argued as Celtic cult sites. The British record has been comprehensively documented but evidence for iron age usage is extremely poor. Stronger evidence for ritual wells and shafts comes from the continent. Shafts have been recovered, for example, inside or under the banks of southern German Viereckschanzen. Bavarian examples occur at Holzhausen, Tomerdingen, Schonfeld and Kreutzpullach. Fellbach-Schmiden (Baden-Wiirttemberg) may also be noted in this context. At Holzhausen, three shafts up to 36.5 m deep were sunk in the Iron Age. The presence of organic material, burning within shafts, and the deliberate placement of a wooden stake and a flesh-hook in two shafts led the excavator to suggest these had been used for the disposal of sacrificial remains. At Fellbach-Schmiden fragments of three deer figurines were recovered from an oaklined shaft (by dendrochronology to 123 BC).

Beyond southern Germany, shafts with iron age fills are rare. Gallic examples are mainly restricted to the Provincia. The fills of some Toulouse shafts date to the first century BC (e.g. the ceramic-filled shaft from Vieille-Toulouse), but none securely predates the Roman intervention. Pre-conquest fills occur elsewhere in the Provincia, as at Nimes, but these shafts seem to have had utilitarian roles. Most Gallic examples postdate the Augustan era, as at Argentomagus  where shafts date to the first century AD and the first- to second-century AD examples at Chartres (Eure-etLoir) Given this dating, it is important to recall that the use of shafts for cult purposes was common in the Graeco-Roman world and it seems very possible that wells and shafts are essentially post-conquest cult loci.

Celto-Ligurian Sanctuaries

These well-documented stone-built structures of Provence include Entremont, Roquepertuse  and Glanum.

They comprise monumental propylaea, decorated with sculptured reliefs, frequently of the human head, and often associated with free-standing sculptures of cross-legged anthropomorphic figures. The dating of some examples is uncertain, but Entremont and Roquepertuse were probably destroyed during the Roman intervention in c. 125 Be. These sites provide much information on sanctuary structure, but they represent an exceptional, localized development, owing much to the long Greek contacts in the Bouches-du-Rhone area, and, with the exception of a dieu-accroupi from Argentomagus , do not appear to be replicated elsewhere in Celtic Europe.

Enclosures and their related structures

Until recently, it was thought that prior to the second century Be, iron age Celtic peoples rarely constructed special cult sites, and possible early sites like the early third-century Be rectilinear enclosure at Libenice (Prague) were considered exceptions to this rule. Excavation in the last decade, particularly in northern France, has changed this position somewhat (the sanctuary at Gournay-sur-Aronde (Oise), for example, has fourth-century Be origins), but it remains true that the majority of excavated cult sites employing enclosures do not predate 250 Be.

Viereckschanzen

This term was originally applied to a series of Bavarian rectilinear enclosures delimited by an earth bank and ditch, enclosing an area of approximately I ha.  Viereckschanzen were classified on morphological and typological criteria by Schwarz, who was influential in arguing for the cult function of such sites. In contrast to the 'Belgic' sanctuaries discussed below, the assemblages from excavated Viereckschanzen are generally very poor, and the cult function of most examples is uncertain. Sites across Celtic Europe, from Gosbecks in Britain to Msecke Zehrovice in Bohemia have been related to the Bavarian Viereckschanzen on morphological grounds. There has been a marked tendency to assume commonality of function across the series, and taking Holzhausen as an indicator, this role has often been argued to be a cult one. The recent doubts cast over Holzhausen are thus significant, not least because, despite numerous excavations, Holzhausen remains the sole site with a strong claim to a cult function. The excavators of Viereckschanzen-comparables in France, for example, have suggested utilitarian roles for enclosures at  and St-Arnoult. In addition, the perceived morphological and topographical similarities between the Bavarian sites and a set of proposed Viereckschanzen-comparables in the Limousin, France, have recently been shown to be statistically non-significant. It is likely that some sites now placed in the Viereckschanzen class served a cult function, but it seems clear that the category subsumes a variety of monument groups. Without independent excavation criteria, as yet lacking, unexcavated sites, and the majority of excavated examples, cannot be seen as cult loci.

Last decade, a series of excavations in the territory of the Bellovaci and Ambiani (Picardy, France) has substantially increased our knowledge of sanctuaries and attitudes to sacred space in Belgic Gaul. Sanctuary sites with features similar to those of Picardy occur beyond Belgic territory, but given the extent of excavation in Belgic Gaul, it is proposed to concentrate on that area here. Belgic sanctuaries include Gournay-sur-Aronde), Estrees-St-Denis, St-Maur  and Vendeuil-Caply  in Dept Oise; Ribemont-sur-Ancre, Morviller-St-Saturnin  and Chilly  in Dept Somme; and Mouzon  in Dept Ardennes. One of the best-understood sequences is that for the sanctuary at the centre of the oppidum of Gournay-sur-Aronde (Oise). The sanctuary site forms a rectilinear enclosure which in its first (fourth century Bc) phase measured 45 x 38 m and was defined by a ditch and low bank. In the late fourth to mid-third century Bc a palisade was added on the external edge of the ditch and in the late third to early second century Be a second ditch was added beyond the first. From its construction in the fourth century Be until C.30 Bc, the primary ditch was used to deposit over 2,000 broken weapons and 3,000 animal bones. At the center of the enclosure a group of pits was dug in the mid-third century Be. Nine roughly circular pits 1.2 m deep, surrounded a tenth, oval, pit in which cattle remains were allowed to decompose.

Around the end of the third century Be these pits served as foundations for a wooden building, replaced by two subsequent first-century Be structures. The latter, constructed c.30 Be, was of [anum form. Ditches serving as repositories for weapons, tools, and animal and human bone, associated with a central arrangement of pits which later saw structural formalization, occur at other Oise sites. These include St-Maur  and Vendeuil-Caply. At St-Maur, ritual activity commenced later than at Gournay, in c.250 Be. Many variations on the Gournay pattern are known, even within Belgic Gaul. At some sites assemblages are characterized less by weapons than by jewellery; examples here are Estn!es-St-Denis, Vendeuil-Caply and Morviller-St-Saturnin. Equally, certain sites, such as Estrees-St-Denis and Montmartin, are less clearly demarcated from their associated settlements than are others in the Belgic group.

A number of sites produce a high proportion of human bone. Among these is Ribemont-sur-Ancre (Somme), 50 km from Gournay. An iron age cult area was discovered here, near the principal temple of a vast Gallo-Roman cult complex. Over 200 long bones (mainly human, but also of horses) had been used to form a box-like construction, r.65 m sq, surrounding a post-hole filled with human ashes. Originally standing 1 m high, this ossuary was surrounded by iron weapons. A second ossuary has recently been excavated. Both examples stood in corners of a ditch-and-bank quadrangular enclosure, itself set in a vast enclosure over 200 m in length. Weapons and bones had been placed along the edges of the former. Two distinct dating episodes are noted at Ribemont: the construction of the ossuaries C200 Be, and a phase of deposition within the ditch, centred on the period of the Gallic War. The specific function of the site is unclear. The bone constructions could represent the massacre of prisoners, the remains of enemies killed in battle, or a collective sepulchre for the 'friendly' dead.

A second site yielding a high proportion of human remains is Mceuvres, Cambrais. Although Belgic sanctuaries thus exhibit considerable variety, two broad shared characteristics emerge: a rich assemblage of deposited material - weapons, often 'sacrificed' by deliberate damage, animal and human bones, and jewellery - and a palisaded/ditched enclosure. Many of the more fully excavated sites have also produced internal structures, but these generally postdate the enclosing works. This point is reconsidered below. Finally, whilst we know almost nothing of deity worship within these loci (or even whether this occurred), it is clear that deposition was a key ritual activity at such sites. These depositional acts were the culmination of a series of ritual processes, including the dismemberment and exposure of human and animal bone and the ritual sacrifice of weapons, well documented at Gournay-sur-Aronde (Brunaux and Rapin 1988) and Ribemont-sur- Ancre.

Throughout the Celtic world, there are very numerous occurrences of preconquest horizons at Roman period lana. The structural associations of these levels are unfortunately often uncertain, obliterated by the subsequent buildings, but frequently comprise pits or pit groupings, similar to those of the Belgic area. A recently excavated example occurs at Vertault (Cote d'Or). Enclosures with affinities to those of Picardy include Mirebeau-sur-Beze (Cote d'Or), Tronoen (Finistere) and Lousonna-Vidy (Switzerland). Weapons assemblages recently reinterpreted as deriving from cult loci include Nanteuil-sur-Aisne and Roizy (Ardennes) in the territory of the Gallic Remi. In Britain, numerous enclosed loci are proposed as iron age sanctuaries. Among these are a square palisaded enclosure surrounding a circular structure at Hayling Island, and a square, multi-ditched later iron age enclosure, overlain by a circular temple of the first century AD, at Gosbecks. Other examples are Uley, Lancing Ring  and Harlow.

The Importance of Enclosure

Enclosure was the primary and indispensable feature of Celtic cult sites. Some general points may be offered on enclosures in Celtic religious contexts. Enclosure is an act of boundary. Prior to the Late Iron Age, this boundedness was often underlined by the physical isolation of cult loci from settlement sites. The early La Tene quadrangular enclosure at the summit of the oppidum of Zavist (Bohemia) is one of earliest proposed cult sites to exhibit close settlement links. By the later Iron Age, many cult loci occur within settlement contexts. Others, like Ribemont, comprise rural loci which under the Roman hegemony develop into major cult complexes.

As the Belgic examples show, the early phases at most cult enclosures, delimited by a ditch and sometimes a palisade, had no accompanying internal structures. Crucially, in many cases, enclosure ditches were themselves a focus of ritual activity, serving as depositional zones. In the Celtic world, an enclosure ditch was often not simply a delimiter of sacred space; it was itself a primary focus of cult activity.

Weaponry, in particular, tends to be associated with ditches rather than with formal structures. The weapons assemblages of both Gournay and St-Maur were associated with the enclosure ditches; those of Ribemont with the ossuaries and the ditch edge, and during the first century BC weapons were also placed in the ditch. Formal structures occur at late stages in the history of these sites. Others may have had no structural phase at all, for example Mceuvres and possibly also the poorly documented closed deposits of weapons and horse gear such as the Masse at Tiefenau and Naillers.

Ditches, as depositional loci, were thus often intimately associated with cult activity. The relative chronology of enclosed cult loci merits stressing in this context. The perimeter - ditch and or palisade - tends to predate internal features. Later structures, often formalizing pit groups (as at Gournay and Vendeuil-Caply) tend to be square or rectangular in plan, culminating in square-plan that relates the form of internal structures to the plan of the enclosure itself, again emphasizing the pre-eminence of the enclosing works. Whether the development of formal structures is due to Mediterranean influence, as some commentators maintain is open to debate. But it is clear that an essential feature of many Celtic cult loci is the act of enclosure, less as sacred delimiter than as itself sacred space.

Space and Symbolism

Our understanding of the beliefs which structured the organization of sacred space among Celtic peoples is very uncertain. But the characteristics of cult sites allow us to see some underlying principles at work, if not the reasons for them. For example, many Celtic cult sites have east-facing entrances. This feature is by no means confined to sacred sites, but the frequent occurrence of easterly orientation suggests it was a symbolic referent informing spatial delimitation in many contexts. Posidonius, who says the Keltoi revere the gods by turning to the right, also recognized the symbolic importance of orientation, and may be describing a rite of circumambulation. In the first century AD Pliny noted Gallic circumambulation, apparently in the opposite direction. Similar rites have tentatively been evoked to explain the presence of an annular space, delimited by two palisades, at the St-Maur sanctuary. Other astronomical considerations may have influenced the structuring of space within enclosures. The internal structures at Libenice (Prague) were arranged with reference to the position of the sun and the solstices, and at Gournay (Oise) four posts at the centre of the enclosure marked the cardinal points.  Caesar's comment  on the druidic meeting-place at the center of Gaul may suggest the symbolic importance of centrality to Celtic peoples. This may also be reflected in the widespread occurrence of the element medio- (center, middle) in Celtic place-names . The British Medionemeton  is particularly relevant in this context. It is less clear whether this concept informed the structuring of sacred spaces on a microcosmic level, although Brunaux comments that most temples are placed at the centre of enclosures. Brunaux has also argued for a geography in which every natural phenomenon was deemed sacred by the Celts. At present, however, ideas on Celtic sacred geography remain speculative, articulated mainly by appeals to the medieval literature.

(The Celtic World, 1995, Jane Webster)