Botunical Journal of the Linnean Society, 73: 177-185. With 1 figure

July/September/October 1976

Observations on the bracken component of the pre-Hadrianic deposits at Vindolanda, Northumberland M. R. D. SEAWARD, F.L.S.

School o f Environmental Science University of Bradford, Bradford

Recent excavations at Vindolanda have disclosed that the floors of pre-Hadrianic buildings were mainly composed of Pteridium, with considerable quantities of straw, several species of trees, and some bryophytes, notably Hylocomium splendens. The thickness of the bracken deposits suggests that bracken-hawesting was a major occupation of the community, which may have used it for winter bedding. Conditions apparently were very insanitary as the ‘bedding’ of a 30 m2 area contained fmillion puparia of the stable-fly (Stomoxys culcitmns). The bracken deposits were encapsuiated between layers of clay in anaerobic b u t not, as far as is known, water-logged conditions, and together with urine impregnation provide a remarkable medium for the preservation of natural and man-made materials, including a unique collection of writing tablets which give information about Roman Britain around AD 100.

Introduction . . . . . . . Location and origin of Vindolanda Occupational levels . . . . . Analysis and origin of floor materials Preservative qualities of bracken . Conclusion . . . . . . . Acknowledgements . . . . . References . . . . . . .

CONTENTS

. . . . . . . . . . . . . . . 177

. . . . . . . . . . . . . . . 178

. . . . . . . . . . . . . . . 178

. . . . . . . . . . . . . . . 180

. . . . . . . . . . . . . . . 183

. . . . . . . . . . . . . . . 184

. . . . . . . . . . . . . . . 185

. . . . . . . . . . . . . . . 185

INTRODUCTION

Vindolanda (Chesterholm) was known to be the site of a Roman fort in the days of Elizabeth I. The first serious excavations were undertaken between 1814 and 1835 by the Rev. A. Hedley, but his findings relating to the late stone fort (mainly 4th century) were never published. The modern research programme began in 1930 when Professor E. Birley made a renewed examination of the fort and, by trenching in the vicus, demonstrated the existence of 1st century occupation (Birley, 1931). Other vicus structures were examined on a relatively small scale by Mr R. Birley in 1959 and between 1967 and 1969 (Birley, 1970). The site passed into the hands of the Vindolanda Trust in 1970 and the current large-scale excavations under the direction of Mr

177

178 M. R. D. SEAWARD

R. Birley commenced in 1971. The work of the past four years has shown that the complete excavation of the site is going to be a very long task. Further details of this work are to be found in Birley (1976), and a review of the environmental archaeology is provided by Seaward (1976).

LOCATION AND ORIGIN OF VINDOLANDA

Vindolanda (grid ref: 35/771661) is in Northumberland on a small level plateau (altitude 160-165 m) 1.5 km south of Hadrian’s Wall, above the junction of the Bradley and Brackie’s Burns, with the deep valley of the Doe Sike to the south, giving excellent drainage on three sides and a good water supply. Barcombe Hill, with its sandstone quarries, towers above the plateau to the east, and the ground rises sharply to the north. The Stanegate road, constructed c. AD 80, which connected Carlisle to Corbridge, crossed the burns below the plateau at an altitude of c. 150 m, and the Roman plans required forts every seven miles or so (c. 11 km) along the road; to the west lies Carvoran and to the east lies Newbrough.

When Hadrian’s Wall was constructed (work is thought to have commenced soon after AD 120), the majority of the Stanegate forts were abandoned. Vindolanda lay deserted for nearly 40 years, but by AD 160 it had been re-occupied; and later a substantial civilian settlement (vicus) grew up outside the fort walls, and there the soldiers’ wives and children, retired soldiers, merchants, craftsmen and farmers made their homes. The remains of Vindolanda cover an area of about 7 ha (c. 18 acres), with a fort of 1.4 ha (3 .5 acres) and a vicus of at least 4 ha (10 acres), and there are in places six superimposed structures, with Roman occupational debris to a depth of over 3.5 m. T o the east of the site lies the late stone fort platform in the characteristic ‘playing-card’ shape, and to the west lies the vicus. The pre-Hadrianic area under present consideration lies to the south of the site between the exposed vicus and the visible fort.

Pre-Hadrianic buildings (c. AD 90-125) outside the conjectured area were located by chance while preparing a drain in late 1972. Not until the new digging season in 1973 got under way did the full potential of this chance find become apparent.

OCCUPATIONAL LEVELS

Within the occupational debris uncovered at Vindolanda were two distinct layers, which have since been dated asaAD 90-95 and AD 95-105 (or 110 at the latest)- strata I and I1 respectively in Fig. 1-each of which is characteriz,ed by a very high organic content. Two further occupational debris levels (strata I11 and IV in Fig. 1) are superimposed on these layers and can be dated between c. AD 105 and 125. The partition walls of the pre-Hadrianic buildings are composed essentially of birch uprights (c. 8 cm in diam.) interspersed with seasoned and fashioned oak; the whole is interwoven mainly with birch branches, although other pieces, usually hazel, oak and willow, are frequently found in these strata.

I t will be seen that the occupational debris levels are sandwiched between compacted clay levels (C strata in Fig. l), an indication that the floor of the

BRACKEN IN PRE-HADRIANIC DEPOSITS 179

particular building had degenerated to such an extent that a new and more permanent floor was the answer to the problems (many of them due no doubt to poor sanitation). This new floor may possibly have coincided with partial or total reconstruction and/or replacement of the superstructure. It is the clay compactions which have created the pockets of anaerobic and chemical conditions (see below) that have preserved such a wide and varied range of both natural and man-made materials representative of the period.

Figure 1 . Schematic soil profile to show relative positions and thicknesses of pre-Hadrianic occupational levels (I-IV) at Vindolanda. A, Undisturbed, original sandy-clay layer; B, disturbed sandy-clay layer with high percentage of bones and wood fragments; C, compacted clay; D, rubble and mixed soil in-fill; E, undifferentiated grey clay; F, G and H, occupational levels of the 3rd and 4th centuries.

Around AD 125 an earth and rubble in-fill (stratum D in Fig. 1) brought the occupation at this particular site to a close. A clay soil, lacking stratification (stratum E), is followed by redevelopment of the site in the late 2nd century (stratum F) which continued through into the 4th century (stratum H); a flagged roadway on a par with the junction of strata F and G can be dated to c. AD 270.

Detailed pollen analyses for the Vindolanda deposits and the adjacent soils have yet to be undertaken; interesting data however are available from Muckle Moss, 3 km ENE of Vindolanda (Pearson, 1960) which show to good effect the vegetational changes before and after the advent of Roman occupation in the area. Pearson has adopted the designation ‘VIII modern’ for a pollen zone which begins at a depth of 150 cm and continues to the present surface. Conway (1947) dates this demarcation at AD 1100 for Ringinglow bog in the southern

180 M. R . D. SEAWARD

Pennines, and Pearson gives a date of AD 1000 for Muckle Moss. A noticeable change also occurs at a depth of 270 cm wirhin the VIII zone, where an increase in spores of Pteridium and pollen of ruderals such as Plantago spp- and certain Compositae corresponds to a date of AD 110 according to Pearson. The occasional grains of cereal crops are to be found to a depth of 270 cm, but their presence is much more dramatic after AD 1000. However, it should be noted that the dates suggested by both Conway (1947) and Pearson (1960) are estimates based on the assumption of uniform peat growth; they have not yet been confirmed by C-14 dating.

The change in the flora of the mineral ground for the Vindolanda area may thus be interpreted as follows: (1) the development first of more open woodland conditions of mixed Quercus, with considerable Betula, and probably interspersed with thickets of Corylus and Alnus in the gullies (following a climatic change c. 500 BC) favouring the spread of Pteridium; (2) the continuing forest clearance, and the establishment of dense stands of Pteridiurn; and ( 3 ) the appearance and spread of ruderals associated with the Roman occupation.

ANALYSIS AND ORIGIN OF FLOOR MATERIALS

The organic material of the pre-Hadrianic deposits proved to be composed mainly of Pteridium-the fronds being easily recognizable from the mode of branching, and the sori running all round the margins of the pinnules (the latter indicating a collection date of July or August). Further constituents of the deposits included considerable quantities of straw (the junctions between this and the bracken providing the most suitable niche for the preservation of stable-fly puparia-see below), twigs of ash, aspen (?), hazel, oak, pine, rowan and willow (cf. tree pollen spectrum in zone VIII of Muckle Moss analysis-Pearson, 1960) and bryophytes. The last named warrant further investigation, but the most abundant species (forming over 5 5% of the material analysed to date) is Hylocornium splendens, which is among the most commonly encountered bryophytes in British Late Pleistocene deposits including archaeological sites of the Bronze and Iron Ages and of Roman times (Dickson, 197 3). The species is easily recognizable in the Vindolanda deposits due to the good branching material which is several centimetres long in most cases. The large quantities of H. splendens in the pre-Hadrianic deposits leave one to speculate that the moss had an economic value (?as packing or stuffing material), for although it would be common in the vicinity of the settlement, it is hardly likely to have been accidentally cropped in such bulk during bracken-harvesting, especially as it is not a normal associate of the Pteridietum. It is more commonly found on acid and peaty soils amongst grass and heather or in open woodland such as birch; some of this moss may well have been picked up during the collection of birch for the construction of building partition walls.

The next commonest moss encountered is Acrocladium cuspidaturn, a moss with a wide tolerance of soil reaction; in all probability this moss was accidentally picked up through (1) foraging (perhaps for hazel nuts) over the richer soils to be found in the gullies which supported Corylus and A h u s thickets, and/or (2) the collection of clay soils for compacting as floors over

BRACKEN IN PRE-HADRIANIC DEPOSITS 181

occupational debris (C strata in Fig. 1). Further bryophytes so far determined from the pre-Hadrianic deposits are: Brachythecium rutabulum, Pleurozium schreberi, Rhytidiadelphus squarrosus, Thuidium tamariscinum, Polytrichum commune, Mnium undulatum and M. cf. affine (a moss of wetlands or moist soil woodlands which, according to Dickson (1973), has been doubtfully found in British Pleistocene deposits-records should be referred to M. rugicum; the Vindolanda material is not referable to the latter, and ratification awaits the collection of further material).

Other macrophytic remains of the deposits include hazel nuts, acorns, gorse pods, heather stems, the puff-balls Bovista nigrescens and Calvatia caelata, and as yet very limited remains of unidentified leaves and stems. The greenness of the bracken, heather and gorse material on immediate exposure has been erroneously described in the press, for although chlorophyll is to be seen through critical examinations of these plant remains, the phenomenon emotively referred to is actually attributable to the considerable quantity of bryophytes which have retained their olive to light green coloration.

From the thickness of the pre-Hadrianic deposits it would appear that bracken-harvesting was a major occupation of the community at Vindolanda, for one small area of rooms so far excavated (c. 30 m 2 ) would have required at least 1 ha (2.5 acres) of well-cropped Pteridietum. This measurement has been derived from a calculation involving a determination of the volume and dry weight of the deposits and the magnitude of the bracken contained therein, and a knowledge of present-day frond productivity and spatial distribution in a Pteridetum determined from analyses I have undertaken on lowland heaths in North Lincolnshire (Seaward, 1973) and upland moorlands in Yorkshire, and from numerous literature sources (e.g. Pearsall & Gorham, 1956). It is interesting to consider here the possible uses to which bracken was put at Vindolanda, during a period when so little is known of the plant’s economic importance-although Virgil (in Georgia, Book 111, lines 297-9) recommends its use as winter bedding material for penned sheep-and to relate these to pre- and post-Roman practices in Britain (see Rymer, 1976); for example, bracken is still used as bedding material in some byres of northern England. There are also strong reasons to support the view that the spreading of bracken on the Vindolanda floors functioned to some extent as a sponge for the absorption of water at times (mainly or solely seasonal) of high water-table levels.

There is some evidence to support the theory that the bracken was stored before use, and that layer upon layer was built up within particular rooms at Vindolanda over many months without periodic ‘mucking out’. The bracken layers also contained considerable quantities of straw and other debris, including some domestic excreta (probably human)-the whole being impreg- nated with urine (probably widely used for tanning and not necessarily confined to a single workshop). Originally this medium proved to be a most suitable environment for the development of the larvae of the stable-fly (Stomoxys calcitrans L.). The puparia of this fly were found in abundance throughout the organic deposits-the majority being empty but quite a few containing the fully-developed imago (one could consider here the insecticidal quality of the bracken). Productivity measurements showed there t o be well in excess of f million puparia within the deposits of the 30 m2 area of rooms studied. The standard of hygiene can be well imagined, with the adult fly

13

182 M. R. D. SEAWARD

sucking the blood of warm-blooded animals such as cattle, horses, poultry and occasionally man.

Other arthropods were extracted from the organic deposits at York University following the standard procedures developed by the York Archaeological Trust, whereby the animal remains are floated out of the fragmented plant debris by agitation in a finesjet of water, after mixing the material with paraffin. In addition, a careful visual inspection was made of a large number of intact fronds.

Entomologically, the results were extremely disappointing; the bracken had obviously been cut fairly late in the season, given the state of the expanded pinnae (see above), and a reasonable faunal diversity might have been expected. Of the species found, however, only one (Dusyneura filicina (Kieff)) is a normal bracken herbivore, and even here, only four galls were found despite the large number of frond fragments searched. With one other possible exception, all the other arthropods were almost certainly associated with the debris on the floor of the building, rather than being brought in on the bracken. They included the beetle Anthicus yuisquilius Thomson, a very common species in vegetable refuse, numerous dipterous pupae, and at least three common species of Cryptostigmata (oribatid mites). One of these appears to be identical with a Chamobates sp. found on living bracken fronds at the present time, but because of the problems involved in specific identification of this group it is by no means certain that they are the same species. However, this is certainly not impossible, particularly in view of the fact that the Chamobates sp. found on the fronds may, in any case, be more typically a member of the litter fauna which ‘wanders up’ on to the fronds (see Lawton, 1976: Appendix 1).

The apparent scarcity of ‘normal’ bracken arthropods in the Vindolanda prompted a simple experiment a t York ,University, in which a sample of bracken was cut from Skipwith Common, Yorkshire (see Lawton, 1976) on 1 3 September 1973, and divided into two halves. One half was sorted immediately to extract arthropods, and the other half was spread on the tiled floor of an outhouse and walked on and disturbed regularly for a minimum of five minutes at least once every day for three weeks; this material was also then sorted. Dusyneura galls remained readily recognizable. The approximate recovery rates of the other, free-living species are listed in Table 1. It is clear that recovery rates tended to be very low with the exception of those of the mite Chumobates, which showed every sign of having survived, and possibly even reproduced, during the experiment. This makes it more likely that the specimens from the Vindolanda material may be the same species that is found on the living fronds, and that they alone were able to survive once the fronds were cut and spread on the floor.

Furthermore, as the bracken dried out and was trampled on in the above experiment, it filled the whole room with a very unpleasant smell. It seems unlikely that the Romans would have tolerated this, and it is therefore likely that the bracken was air-dried before it was spread on the floor. If this was the case, it would undoubtedly have led to an even greater loss of animals, and it is hardly surprising that so few typical members of the bracken fauna were discovered.

I t may be concluded from the York University work that Dasyneura filicina,

BRACKEN IN PRE-HADRIANIC DEPOSITS 183

and perhaps the mite (Charnobates sp.), have been associated with bracken for at least 1900 years. There is no reason to suspect that the apparent absence of species other than these reflects anything but the treatment of the bracken prior to its being spread on the floor.

Table 1 . Recovery rates for free-living bracken arthropods after three weeks on the floor of an outhouse-an approximate simulation of the conditions on the floor of a Vindolanda room. (Experimentcarried out by Dr J . H. Lawton at York University; see Lawton, 1976, for further information on the species listed.)

Estimated numbers Recovery rate, initially in material three weeks

Species spread on floor later (%)

Macrosip h urn p tericolens 4000 2.8 Monalocons filicis 120 1.1 Criomorphus pteridis 62 U Charnobates sp. 46 46.4 Bourletiella viridescens 42 0 Sawfly caterpillars 6 0

The pre-Hadrianic deposits also contain a large quantity of molluscs, including oysters, mussels and snails, and considerable quantities of bone, in- cluding cattle (Celtic Shorthorn variety), sheep, goat, pig, boar (?), red and roe deer, horse, dog and a diversity of birds of varying sizes (including game and poultry). These bones, together with their unused adjoining tissues, and probably offal, would have been lying around and presenting a somewhat unpleasant environment; the availability of this material to the carrion bird feeders may be reflected in the large number of their feathers found.

PRESERVATIVE QUALITIES OF BRACKEN

The predominantly bracken deposits (strata I and I1 in Fig. 1) and the straw and occupational debris deposits (mainly 111, and parts of I, I1 and IV), between the compacted clay strata (C, and to some extent D) of the pre-Hadrianic period are encapsulated in anaerobic but not water-logged conditions, and together with the urine impregnation provide a remarkable medium for the preservation of a range of natural and man-made materials.

The chemistry of this medium warrants further investigation. I t would appear that the preservative quality of the medium is connected with the large amounts of organic material; the production of ‘tannins’, on the death of plant cells, is anti-bacterial. Further tannins are contributed in this instance by the considerable quantity of leather material (see below) present in these deposits, and there is the further possibility that tannins may have been purposely produced on these, and other, premises for leather-work, as is indicated by the extensive number of off-cuts present, and the urine-impregnated medium.

Vivianite, a form of iron phosphate, was much in evidence; the enormous quantity of bones present in the deposits would contribute to phosphate

M. R. D. SEAWARD 184

availability. The pH measurements of the organic deposits and occupational debris from all the pre-Hadrianic levels so far studied were remarkable in their homogeneity, all being within the limited range of 5.3-5 -6.

The medium has preserved a unique collection of writing tablets (nearly 300 fragments had been discovered by the end of the 1974 digging season), which not only supplies evidence for a period of Roman Britain hitherto little known but also gives a background knowledge of everyday life through accounts, quarter-masters’ lists, receipts, letters, etc. The main form of writing tablets are thin slivers cut from fine-grained deciduous and coniferous wood (the former probably includng lime (Tiliu), and therefore imported since this genus appears to have been rare in this region at that time according to pollen analysis-cf. Pearson, 1960; Chapman, 1964). The tablets have proved most useful in the confirmation of the dating of the pre-Hadrianic deposits as proposed by Mr R. Birley; one of the tablets, for example, contains a rererence to a man who can tentatively be identified as Lucius Neratius Marcellus, known to have been governor of Britain a t the beginning of AD 103. Further details of the discovery, state of preservation, treatment and general nature of the Vindo- landa writing tablets are given in Bowman & Thomas (1974). Also preserved in the pre-Hadrianic organic deposits are considerable quantities of leather (the 1973 excavation alone revealed nearly 1000 items) together with some of the leatherworkers’ tools, and pieces of cloth which vary in size and quality of weave and pattern.

CONCLUSION

Two major factors need consideration in the interpretation of the nature and content of the pre-Hadrianic deposits at Vindolanda: the first concerns the chemical complexity of the bracken, and the second concerns the chemical complexity of the organic deposits in toto. As has been seen elsewhere in this issue (Cooper-Driver, 1976; Evans, I. A., 1976; Evans, W. C., 1976), bracken has an interesting chemistry and contains a wide range of secondary plant products. Its chemistry varies with its development, and it is worth considering the stage at which the harvested bracken was utilized on the Vindolanda floors. As has been mentioned earlier, a floor strewn with freshly-cut bracken and bruised through trampling would create an unpleasant environment in which to live and work, but at the same time it would produce in all probability an effective insecticidal exudate. The latter would certainly be necessary judging by the number of stable-flies preserved in the deposits, and its effectiveness is shown by the suppressed imago development in many puparia. On the other hand, material which had been stored and dried correctly would provide a much more favourable environment in which to live and work; furthermore, the tannins produced by such treated bracken may have been utilized in leather-working on the site, or would certainly have been directly available in the deposits to aid preservation of the extensive man-made and natural materials, a few of which are enumerated above.

The chemical composition of the organic deposits in toto and its role in the preservation of these materials is even more complex; bracken may contribute much or little to this phenomenon. It remains to be seen if bracken is a common constituent of further pre-Hadrianic exposures at Vindolanda, and

BRACKEN IN PRE-HADRIANIC DEPOSITS 185

how the knowledge gained from this work can be applied to other comparable archaeological sites.

ACKNOWLEDGEMENTS

I wish to record my thanks to Mr and Mrs R. Birley for their most valuable help in the preparation on this paper and for their 'criticism of the draft manuscript, to Dr J . H. Lawton and Mr G. Hodgson for providing me with unpublished data, and to Mrs E. Broadhead, Miss V. A. Hinton, Dr J. Turner and Dr R. Watling for their valuable criticism of various aspects of this work.

REFERENCES

BIRLEY, E., 1931. Excavations at Chesterholm-Vindolanda. Arch. Ael. (3rd ser.), 8: 182-212. BIRLEY, R., 1970. Excavations at Chesterholm-Vindolanda, 1967-1969. Arch. Ael. (4th ser.), 48 :

BIRLEY, R., 1976. Vindolanda. London: Thames & Hudson (in press). BOWMAN, A. K. & THOMAS, J . D., 1974. The Vindolanda writing tablets. Newcastle upon Tyne: Frank

CHAPMAN, S. B., 1964. The ecology of Coom Rigg Moss, Northumberland. I. Stratigraphy and present

CONWAY, V. M., 1947. Ringinglow Bog, near Sheffield. Part I. Historical. J. EcoL, 34 : 149-81. COOPER-DRIVER, G., 1976. Chemotaxonomy and phytochemical ecology of bracken. Bot. J. Linn.

DICKSON, J . H., 1973. Bryophytes of the Pleistocene. Cambridge: Cambridge University Press. EVANS, I. A,, 1976. Relationships between bracken and cancer. Bot. J. Linn. SOC.. 73: 105-12. EVANS, W. C., 1976. Bracken thiaminase mediated neurotoxic syndromes. Bot. J. Linn. SOC., 73:

LAWTON, J. H., 1976. The structure of the arthropod community in bracken. Bot. J. Linn. SOC., 73:

PEARSALL,'W. H. & GORHAM, E., 1956. Production ecology. I . Standing crops of natural vegetation.

PEARSON, M. C., 1960. Muckle MOSS. Northumberland. I. Historical. J. Ecol., 48 : 647-66. RYMER, L., 1976. The history and ethnobotany of bracken. Bot. J. Linn. SOC., 73: 151-76. SEAWARD, M. R. D., 1973. The ecology of Scunthorpe heathlands with particular reference to

SEAWARD, M. R. D., 1976. The Vindotanda environment Haltwhistle: Barcombe Publications.

97-155.

Graham.

vegetation. J. Ecol., 52: 299-3 1 3 .

SOC., 73: 35-46.

113-31.

187-216.

Oikos, 7: 193-201.

Twigmoor Warren. J. Scunthorpe Mus. SOC. (Ser. 2) (Nat. Hist.), No. 2.