["The Archeology of the Roman Economy". by K. Greene, pp. 67-94. 1986 by the University of
California Press.]

Almost everywhere a large part of the population was engaged in agnculture at a relatively low level, while industry depended on a backward technology and was rarely organised in large units.

Ancient historians are unanimous in accepting the importance of agriculture in the Roman empire, to the extent of contending that it was so important that no other form of economic activity could possibly be considered as sign)ficant. On the other hand, the Roman empire shows ample evidence for military power, sophisticated communications, imposing cities, lavish rural villas, metallic currency, and a wide range of goods such as pottery or metalwork which were traded over considerable distances. Are these phenomena consistent with an economy centred entirely upon 'low- level' agriculture, or do they suggest that it was in fact sufficiently productive to support a sign)ficant amount of non-agricultural activity? Comparisons with historical and anthropological evidence from medieval and modern times may well be instructive in answering this question. First, however, it is necessary to examine the state of current knowledge about Roman agriculture, using all possible forms of evidence.

Agriculture is perhaps the most complicated aspect of the Roman economy to study. A deceptive impression is given by the fact that some of its elements are extremely well documented in literary sources. Unfortunately, the documented areas (in both the technical and geographical sense) cannot safely be used to generalise about the agricultural economy of the empire as a whole. The role of archaeology is not much clearer; indeed, some of the most important

activities in the agricultural sector, such as the supply of grain to Rome, would be difficult if not impossible to establish from archaeological evidence on its own. Archaeology can do two main things: first, it can help to place agriculture into a general perspective, by examining the form and extent of rural settlement; second, it can give specific insights through the excavation of farming sites, with full attention to environmental evidence such as plant remains, animal bones and soils. The results of both approaches must of course be integrated with the surviving literary accounts of agriculture. Prehistorians may be shocked to find that in the publication of a British Museum symposium on Roman villas in Italy in 1980, it should still have been necessary for Potter to spell out something which has been a commonplace to them for at least 50 years:

. . . villa-excavation is a complex business involving delicate chronological analysis, exploration of associated barns and field systems, and study of material both robust and ephemeral: seeds, bones and coarse-ware sherds have their story to tell as much as mosaics and architecture.

This chapter will examine the ephemeral as well as the robust; plant remains, animal bones and soils can have far-reaching implications for the interpretation of individual sites or regions, general aspects of the economy, or even such fundamental themes as the origin of Mediterranean civilisation and the climatic history of the world. The aim of what follows is to provide a critical perspective on the kind of things some of these archaeological and scientific specialisms can (or, where appropriate, cannot) be expected to achieve. No attempt will be made to present a synthesis and discussion of Roman farming methods based on literary sources, for this service

has been performed in a comprehensive manner by K. D. White in a number of books and articles .

SOURCES FOR THE STUDY OF ROMAN AGRICULTURE

Extensive literature on Roman agriculture survives from a range of authors who wrote between the second century BC and the fourth century AD. It includes actual manuals of agricultural practices by Cato, Varro, Columella and Palladius, and, in addition, Pliny the Elder's Natural History also contains much information relevant to agriculture (White `97O, ~8-3~). The prevailing tone of the writers other than Palladius is one of more or less well-informed advice from one respectable Italian senatorial land-owner to another; Palladius, the latest of the writers in date, gives instructions, many of which are annotated summaries of the earlier writers. As any reader of modern books on gardening will know, they are not necessarily written by the best gardeners, and what is recommended in their pages is not always practical. Thus, valuable though it is to have the luxury of several contemporary manuals on Roman agriculture, their worth must not be overestimated. They were written by (and for) a privileged few, from their own limited experience of some agriculturally favourable parts of Italy, combined with knowledge drawn from earlier Carthaginian and Greek writers (Martin ~97~). Whilst they do demonstrate a range of possible forms of agriculture, they are by no means comprehensive.

A second form of written evidence is provided by inscriptions on stone, which also have the advantage of being original documents, which have not suffered from errors made by the generations of copy-makers who preserved the literary works. In Egypt, an appreciable number of original documents written on papyrus has also survived . Useful evidence for the study of agriculture obtainable from these sources includes price lists and customs charges (Diocletian's Edict); boundary markers and maps of land divisions ; and details of legal arrangements for the restoration of waste land .

No attempt will be made in this chapter to discuss fundamental issues for which archaeology does not provide direct evidence, such as the use of slaves or tenants as a labour force (Kolendo ~ 976; Wiedemann ~ 98 ~ ). The reader may be interested to investigate some questions which have been approached from opposite directions, using literary or archaeological sources; for example, Frayn relied almost entirely upon literature in her account of subsistence farming in Roman Italy , but Applebaum used the plan of an excavated villa in Roman Britain to deduce its economy. Spurr's study of millet cultivation is also entirely literary, and provides an excellent example of the information with which archaeological and botanical evidence should be compared. Very different is a discussion of the beneficial effects of a plant, medicago, in modern Libyan dry farming); it is suggested that it may have been used to restore soil potential in the same way in the Roman period.

Barker has expressed strong personal views about the relationship between literary and arch- aeological evidence, conditioned by the fact that his main area of research has been in prehistoric archaeology: The archaeological data have to be studied in their own terms first, just like prehistoric data, and not simply worked into a historical framework from the outset, with dubious data from one discipline 'ex plained' by similarly dubious data from the other: in this case archaeological data will invariably tell you what you knew already - comfortingly, but quite spuriously, and expensively at that. In his view, the real role of archaeological data is their potential for 'greatly enhancing our uncder standing of major social and economic processes in the classical world'. His discussion of the environmental remains from Benghazi on the Libyan coast includes an example - an evolving series of interactions between the city, rural settlements and nomadic peoples). Before leaving the subject of literary evidence the value of circumstantial details in the referencee books of writers such as Columella or Pliny may be noted. Pliny the Elder tells an old anecdote which is intended to reinforce his views on the importance of well-managed hard work: Gaius Furius Chresimus, a liberated slave, was extremely unpopular because he got much larger returns from a rather small farm than the neigh bourhood obtained from very large estates, and he was supposed to be using magic spells to entice away other people's crops. He was consequently indicted by the curule aedile Spurius Albinus; and as he was afraid he would be found guilty, when the rime came for the tribes to vote their verdict, he brought all his agricultural implements into court and produced his farm servants, sturdy people and also according to Piso's description well looked after and well clad, his iron tools of excellent make, heavy mattocks, ponderous ploughshares, and well-fed oxen. Then he said 'These are my magic spells, citizens, and I am not able to exhibit to you or to produce in court my midnight labours and early risings and my sweat and toil'. This procured his acquittal by an unani mous verdict. (natural History 18.8)
The whole point of the story would be lost if it were unthinkable that Chresimus should own these items of equipment on a small farm; such details are in many ways more interesting than the lists of essential tools and activities prescribed by Columella for the running of an ideal estate.

Roman villas in many parts of the empire were decorated with paintings, or with mosaics, which because they were laid on floors and made ol robust materials have a much higher survival rate than paintings. Many contain ref:rences to agriculture, sometimes by means of symbolic representations of the seasons, or less frequently by detailed scenes offarming activities (PrecheurCanonge rg6r; Dunbabin rg78). Although the distribution of these is very uneven, they certainly bring the subject to life, and illustrate practices for which direct archaeological evidence has not been found, such as harvesting, ploughing and hunting. The tools involved are frequently quite distinct, and thus help in the interpretation of artefacts found in excavations (Rees rg7g; Glodiaru rg77). Tools and other symbols of agriculture are particularly common on tombstones and religious carvings in Phrygia, Asia Minor (Waelkens rg77), whilst small metal models offarming implements have been found in Germany (e.g. Ternes rg76, 958, Abb r8r, from Rodenkirchen). Sculptures from north-eastern Gaul have allowed a composite picture of a Roman reaping machine, the uallus, to be built up; it is described by Pliny the Elder, but no actual fragments have so far been identified (White rg67b; rg84,6r, figs. 47-48; Heinen _976, 8~-gr, Abb 2a-b).

Over 50 years ago, Loeschcke provided an excellent example of the integration of literary, artistic and archaeological evidence in a study of wine production.in the Mosel region (rg3~), an area outside the range of the Roman agricultural writers discussed above. Since Loeschcke's day, the quantity of carvings and inscriptions has remained static, but archaeology has expanded dramatically to become the principal means by which our understanding of agriculture can be increased. It is able to generate new and exciting areas of evidence both on a general scale by regional fieldwork, and on a detailed level through the location and excavation of individual sites such as villas or farmsteads. Furthermore, modern archaeology works hand-in-hand with many scientific studies which can advance our understanding of the environment in which Roman agriculture operated, such as geomorphology, soil science, climatology, palaeobotany and pollen analysis. Unfortunately for the study of the Roman economy, excavations carried out on many sites have been directed towards the study of architectural features; detailed investigations of outbuildings, soils, animal bones and plant remains are still rare. Scientific precision and thoughtful sampling are even rarer; a substantial excavation report on an important villa in Britain contains.the following preface to a specialist section on molluscs: 'Snails from Pit 0, Room 7 (A very large number of snails were found in the pit, and about l oo were sent for examination)'.

Readers of excavation reports should not accept their conclusions without considering the complex range of factors which can distort their validity. For instance, there are many technical difficulties involved in the excavation of rural sites. Insubstantial timber buildings will have decayed rapidly after their abandonment, and their remains are easily removed by ploughing. Masonry structures which underwent expansion and alteration over several centuries require sensitive stratigraphical study if their structural histories are to be dated correctly. The latest phases of their construction may have removed traces of earlier developments, and in areas without plentiful building stone the remains of their walls are likely to have been subject to disturbance by stone-robbing ).

Perhaps the most difficult problem of all is that agricultural sites do not consist of settlements alone: an inseparable element in the understanding of agriculture is the study of field systems and the extent of the lands which they exploited. .The physical structure, amenities and decoration of a villa, or the wealth of artefacts found on the site of farmstead may reflect the prosperity of their ,ahabitants, but will not reveal the basis of that wealth. In Italy, agricultural activities were frequently carried out in sections of residential buildings; for example, olive and grape presses were located next to domestic facilities in the villa at Sette Finestre . In other areas, these functions might be located in subsidiary structures like the animal sheds and olive oil production unit at the farm known as LM4 in Libya). The boundaries of land units such as villa estates are extremely difficult to define without detailed documentary evidence, which rarely survives before the medieval period in most areas of the Roman empire, although attempts have been made using varying amounts of information and guesswork.

See a chart mapping the agricultural uses near Benghazi on the Libyan coast

Techniques of territorial definition and sitecatchment analysis, derived from geography, can be applied to individual sites or distribution patterns of agricultural settlements, but they rely on many untestable assumptions. Essentially, such studies define a theoretical territory around a site, and measure the proportions of different resources inside it. Applications to sites of the Roman period are rare, but Ellison and Harris found that Roman villas in Sussex possessed particularly favourable mixtures of good arable and grazing land. Regional fieldwork studies such as those discussed in Chapter 5 involve many problems of interpretation, particularly where comparisons between different areas are attempted. It is well known that sites will only be revealed under special circumstances, whether by means of aerial photography or field-walking; for example, permanent pasture or afforestation will inhibit both. Thus, it would be possible for two areas whose settlement patterns were identical in the Roman period to appear quite different in archaeological surveys because of varying degrees or types of modern agricultural exploitation. In all cases, results from such surveys should be taken to represent a minimum view of the original extent of settlement. All of these aspects may be assisted by broad scanning techniques such as satellite imagery.

Many of the surveys carried out in Italy have graded sites according to size, architectural complexity, and the range of datable types of pottery found. However, some excavation is essential to estimate the extent to which surface finds provide a reliable indication of the nature, extent and date of a buried site - in fact, results obtained on sites in Libya allow a degree of optimism. A deeply plough-damaged site should provide plentiful surface finds of pottery and building materials, but if a prosperous early site was overlain by a mediocre later one, and damage was only superficial, surface remains would be unlikely to reflect the status of the earlier phase. However, such problems should

not prevent the undertaking of such surveys, or the interpretation of their results, as long as nonarchaeologists are fully aware of these technical factcrs and are suitably cautious in drawing conclusions from this kind of evidence. The problems involved will be discussed further in relation to individual cases in Chapter 5.

Environmental archaeology has long been a well established subiscipline of prehistoric archaeology . Its role has, however, been taken more and more seriously in historical periods, because of a growing appreciation of its potential contribution to understanding the economic background to social and political events. It can provide specific information, such as demonstrating the function of a building; the classic example is the identification of remains of insects which infest grain-stores from a Roman riverside structure in York . On another level, soil science and plant biology can document longterm regional sequences of geomorphology and vegetation which have implications for the effects of climatic change and the impact on the landscape of human action such as deforestation and ploughing.

The three most important agricultural products traded in the Roman world were grain, wine and olive oil; because of their ubiquity around the Mediterranean today, the plants which produced them are sometimes known as the 'Mediterranean triad', and their farming as 'polyculture'. Literary evidence demonstrates the importance of corn supplies to Rome, and the complex arrangements which ensured its transportation from Egypt and Africa; the thousands of pottery amphorae which carried wine and oil found on sites and in shipwrecks are a vivid record of their contents. However, the familiarity of these crops tends to diminish our concept of their sign)ficance. Work in the Near East and Greece by prehistorians has charted the domestication of various plants, and has assessed their sign)ficance for human settlement. The detailed accounts of different plant species which Jane Renfrew has been able to document, using samples from excavated pre- historic sites , underline the further, and as yet under-exploited, potential of botanical studies in the Roman period.

The cultivation of wheat and barley began in the Near East soon after 6000 BC, with the addition of lentils, peas, and flax for oilseed; fruit and nuts were gathered in the middle and late Stone Ages, including wild grapes and olives . The earliest farmers in south-east Europe were tied to permanent settlements by the need to plant, attend and harvest crops - a constraint not imposed by herds of animals. The objective was therefore to provide a storable surplus of large seeds which would provide essential foods throughout the year; diversification of crops helped to guard against failures. Wild grapes and olives grew all around the coasts of the Mediterranean . Cultivation had begun before 3000 BC in Egypt and Syria, and is attested in Greece around 2500 BC. Although processing reduces the food-value of grapes and olives, they can be converted into wine and oil for long-term storage; finds of grape skins, stalks and pips from early Minoan Crete suggest wine-making, and olive presses are also known in Minoan Crete.

There is profound significance in grape and olive cultivation as opposed to the gathering of wild fruits. These plants require different soil conditions from cereals and pulses, and are harvested later than those crops. Thus, an area of mixed land can proluce two extra harvests with the same amount of labour; however, an investment of time is required to plant, grow and tend the vines and trees before they will produce crops. In Colin Renfrew's view, '. . . the development of Mediterranean polyculture was as important for the emergence of civilisation as was irrigation farming in the Near East. He envisaged a spiral of development, as the demands of a growing population led to the exploitation of tree and vine crops, which in turn led to a higher level of production which could support more people, a proportion of whom could be engaged in work other than agriculture. The coincidence of Minoan and Mycenean civilisation with the areas of early olive and grape cultivation in Europe is therefore highly significant.

Graeme Barker has explored the significance of the cultivation of the 'Mediteranean triad' in Italy. In the Molise valley (on the east side of the Appenines), farming settlements gradually exanded from before 4000 BC until after looo BC, until even marginal land was thoroughly exploited. Then, the cultivation of olives and grapes coincided with a new expansion of settlements .associated with changes in the whole structure of the economy . Barker discussed the relevance of two 'models' which might help to explain the observed changes: Renfrew's 'multiplier effect' , and Boserup's ideas on the interaction of population growth and agricultural change . He found a satisfactory answer in a combination of the two. The 'multiplier effect', rather than contacts with Greek cities in the south of Italy, was aufficient to explain growth associated with polyculture in the first millennium BC. The contrast between this episode of transformation and the thousands of years of continuity which preceded it fitted fairly well with Boserup's model of the conservatism of agricultural systems, which only change when population pressures make it absolutely essential. Thus, population had risen tc the extent which demanded change, whilst the form which the change took initiated the 'multiplier effect' and led to further growth.

First in Etruria, then in outlying regions like Molise, we can see how the new society had to be sustained by an agricultural system that brought with it a transformed world of commercial organisation and social differentiation. In an extraordinary reversal of roles the prehistoric societies of central Italy changed in the space of a few centuries from a virtually Stone Age people . . . into the central nations of the Roman world.

Botanical studies are crucial for the reconstruction of the early history of agriculture and the development of civilisation, whilst pollen analysis has a prominent place in the interpretation of the history of vegetation, as well as possible changes in climate. More specifically, the study of plants from individual sites can help to elucidate their economies as well as the general local environment. Samples must be recovered with care from sign)ficant datable deposits, with knowledge of the effects of different soil conditions upon preservation. Very wet or arid conditions favour the survival of actual seeds, stalks, etc, which can also be preserved as charcoal by burning. Such remains can be separated from samples of soil by some form of fine sieving, but pollen has to be isolated by more sophisticated laboratory methods.

The hot and dry conditions caused by the volcanic ash which buried Pompeii in AD 79 have allowed the survival of remarkable remains of fruit, nuts, beans and seeds. Together with the careful excavation of cavities left in the soil by the decay of roots, these remains demonstrate that sizeable areas of Pompeii were used as orchards, vineyards and market gardens. Conversely, it was the permanently waterlogged conditions which preserved plant remains from a site near the Greek colony at Metaponto, in southern Italy (Carter et al. ~ 985); a water-basin at a spring sanctuary even contained offcuts from the pruning of vines and olive trees, whilst pollen analysis detected the introduction of intensive olive growing in the fourth century BC.

In Libya, plant remains and pollen found in dry conditions at Ghirza in the pre-desert included species that demand much more water than can be provided by normal rainfall . They help to confirm the intepretation of walls and cisterns in adjacent valleys as parts of a careful management system for the exploitation and storage of the little rain which did fall . Seeds from various plants including crops and weeds preserved in wet conditions in the Thames valley near Oxford show that lowlying grasslands were more permanently settled in the Roman period than in the preceding Iron Age, and that they were divided up by hedges. Rescue archaeology has considerably increased the number of excavated plant remains; Jones has listed an impressive array of items from grapes and dates to cucumbers and coriander now known from Romano-British towns, many of them imported rather than home-grown. The military fortress at Novaesium (Neuss) on the lower Rhine in Germany has produced a catalogue of plant remains ranging from obvious items such as grain to numerous examples of medicinal herbs. At nearby Xanten, Knorzer compared plant remains from military and civilian contexts, and found that the army had access to a wider range of soft fruits and pulses, which must have contributed to a healthy diet.

Plant remains may have most to offer archeologists in the provinces of the Roman empire away from the Mediterranean where documentation is least adequate. The high degree of sophistication of animal and plant studies in Germany, the Netherlands and Britain allows comparisons to be drawn between prehistoric and Roman finds which are important in the understanding of the impact of the Roman conquest. Jones has discussed the question of change and innovation in British agriculture, which places the Roman period into an interesting perspective. A major period of new crop introductions seems to have taken place, not at the beginning of the Roman occupation, but between 1000 and 500 BC, which was 'the main period of change between the early Neolithic and the 16th century AD. The species involved are particularly suitable for growing on low-lying heavy soils, which began to be cultivated extensively from the early Iron Age, when ploughing was facilitated by the use ot iron-tipped ploughshares. Superior 'eared' ploughs which could turn a furrow came

into use by the early Roman period at the latest more sophisticated ploughs may have been in use in the late Roman period, but direct evidence is scanty; large scythes and sickles were certainly in use.

Jones claims that the pattern of innovation in Romano-British agriculture demonstrates stagnation in the early empire, and investment in agriculture only in the late empire . This is not compatible with a view of conquest and taxation stimulating production, and coin-based markets promoting 'capitalist' agriculture for profit. Whatever the truth of the matter, it can be seen that seeds and grains painstakingly recovered from samples of soil in the laboratory have implications which extend nght to the heart of the economy and politics of the Roman empire. It is no longer sufficient to make lists of crops which appear in Roman literature and on sites; their growing conditions, yields and nutritional value all need careful evaluation. For example, the bread wheat which is universally cultivated today is easy to harvest and produces high yields of grain which separate easily from the ears. However, before the use of weedkillers and fungicides it was much less resistant to infestation and disease than other arieties of wheat such as emmeror spelt.

If the pattern of innovation and crop introductions is significant, so are the economic implications of crop yields. Research into prehistoric farming methods at Butser experimental farm in Hampshire has achieved far greater returns of grain than had previously been considered likely. The economic implications are considerable; even on conservative estimates, Manning has calculated that the military garrison of Wales could have been fed without difficulty from local land, without recourse to the areas of south-eastern England traditionally considered to be the productive 'granary' which fed the 'highland zone'. In a recent review of this and other calculations, Scott took account of the Butser yields, and contrasted alternative high and low estimates for the amount of land required to supply the grain requirements of the army of Roman Britain; the results were over 80,ooo ha (197,680 acres) or under 2000 (4942 acres), depending on which variables were selected. In isolation, such figures tell us little, but they do allow us to make better informed guesses when attempting to produce 'models' of how we think that the Roman economy worked.

The crop-yields discussed above arejust one result to have come from the Butser experimental farm. Many hypotheses about buildings, methods, crops and animals have been tested, with remarkably clear-cut results. A good example relevant to Roman cereal &rming is the case of T-shaped 'corn- drying ovens', a form of structure found on many sites and assumed to have been used for drying grain in advance of storage. A replica based on excavated remains proved totally unsuitable for this purpose, but did turn out to be effective for sprouting grain for malting and beer-brewing (Reynolds and Langley rg7g). Thus, a further form of secondary agricultural production was suggested by the rejection of the original hypothesis. Brewing was significant enough to be commemorated on the gravestone of a woman from Trier in Germany; not all ofthe barrels known from sites and carvings need have contained wine. At thre opposite end of the Roman empire, the experimental farm at Avdat in Israel has demonstrated the effectiveness of water-conservation methods in desert areas .

Studies of animal bones from Roman excavations have begun to demonstrate that they may be as informative in this period as in prehistory, where, like plant remains, they have been used for evidence of husbandry and diet for many decades. Barker has outlined the raults of bone and other environmental evidence which contradict the supposed dominance of southern Italy by stock-raising; in Rome itself, a deposit dated to the fifth century AD has been published, which allows direct comparison with written records for the supply of meat to the city. The bones from the Schola Praeconum included large quantitia of food debris amongst which pork, beef, and to a lesser extent lamb, were attested. A good idea of the balance of meat consumption and the sources from which meat was drawn can be gained from written records. The bones themselves allow further details to be added; for instance, the animals were as small as those found on prehistoric and medieval sites, showing that large 'improved' breeds are a very recent phenomenon. How ever, the narrow age-range during which the I animals had been slaughtered implies that the fattening of the animals was very efficient, and hat the quality of feeding was very high even when Italy was undergoing crises and invasions in the late Roman period (ibid. 88-89).

Many more sites in the north-western provinces of the Roman empire have properly excavated and published bone reports than around the Mediterranean. Large numbers of reports are important if comparisons are to be possible. Teichert has made a quantified examination of the frequent observation that cattle on Roman or 'Romanised' sites in Germany were larger than those on native sites ( ~ 984). He compared a large number of measurements from samples both sides of the Roman frontier, and demonstrated that, although the ranges overlap, cattle within the empire were of a consistently larger average size. However, the existence of some large cattle outside the frontier but near enough for trade implies that Roman stock or methods of husbandry were known in free Germany. The most sign)ficant observation is that the presence of large cattle is no longer attested north of the Alps after the end of the Roman occupation; Teichert concludes that the necessary knowledge of stock-raising did not exist amongst 'natives', or was lost in the confusion of the babarian migrations. An alternative interpretation is that the dis appearance of urban and military markets may have favoured a return to the selection of smaller, more manageable, animals.

In Britain, Maltby has published clear evidence for a complex butchery trade revealed by bones from Rowan towns. At Exeter, Silchester and London, deposits of cattle bones have been excavated which contain disproportionate numbers of bones from the extremities - skulls, jaws, lower leg - which result from the preliminary preparation of carcasses after slaughter. At Cirencester, 'secondary' butchery was indicated by cut and sawn bones found in pits adjacent to a market hall, which presumably contained rubbish from butchers' shops. At Gloucester, 'secondary' jointing of carcasses prepared elsewhere was still practiced in the fourth century AD. Preliminary study of the age and sex of animals indicates specific production for meat rather than the eating of cattle which had already been used for milk production or traction, whose bones would reveal a wider age-range and include older animals. Female cattle are commoner in towns, presumably because they were easie- to drive. Maltby's evidence does not show whether the meat supply had any official organisation, or whether it operated privately; the separation of stages of carcass preparation and butchery do indicate a level of complexity which must be accounted for in any interpretation of the economy. This evidence for the butchery trade is interesting in relation to a carving,from Trier in Germany, which shows a butcher with a selection of joints and knives.

The potential of further combined archaeological excavation and zoological study for understanding of the place of animals in Roman agriculture is immense. If it is to be properly used archaeologists and historians must appreciate the full complexities involved in the interpretation of animal remains, as Luff has outlined recently. For instance, the bones which are found around a Roman villa will probably reflect the cuts of meat which were popular on its dinner table, and have little relevance to its stock-raising economy. Excavation must make the context of any sample of bones clear, and determine if it formed rapidly or over a long period, and whether it was subject to disturbance after deposition; quite large bones can easily be removed and buried elsewhere by dogs.

Fish and molluscs were also exploited and farmed in Roman times, and their bones and shells deserve careful study. The fish bones from Benghazi shed light upon the forms of coastal fishing that were practiced. Bones found inside amphorae can indicate whether they contained salted fish (complete skeletons) or fish sauce (fragments only), rather than wine or-oil. The sizes of the bones can indicate the season and location of fishing, thanks to the known habitats of fish at different stages in their life-cycles. Impressive systems of tanks, dams and channels were used for fish farming on the Mediterranean coast, whilst many remains of fish-salting and sauce manufacturing centres have been recorded on the Atlantic coasts of Morocco, Spain and Brittany. Oyster shells are found in great quantities on inland sites, and emphasise the ability of the transport system to deliver them in fresh condition. Products of the sea and shore must have provided return cargoes for river boats transporting inland goods to ports; Amand has suggested that this was a factor in the supply of building stone from Tournai in Belgium to coastal sites.

Animals provide many useful by-products, such as wool and leather, which only survive in exceptionally wet or dry conditions of preservation. Each has its own industry, from fleece and hid preparation through to woven textiles and leather goods, whilst bone and horn are themselves raw- materials for small items ranging from tools to furniture inlays. Careful scientific study is necessary to establish the species from which these materials were derived, and the study of wool in particular has allowed the evolution and spread through Europe of different colours and qualities of fleeces to be charted. Sheeprearing has been studied in detail by Frayn, whilst textiles are the speciality of Wild, who has written many specialist reports on fragments retrieved from excavations, as well as a monograph. Fascinating details can emerge fram textiles; some fragments from London indicate a mixture of Mediterranean and northern yarns and techniques. The wool trade of the town of Pompeii is the subject of a comprehensive study by Moeller.

Unlike meat, milk products are most unlikely to leave any trace detectable by archaeologists, other than possible cheese-making equipment such as pottery strainers or presses, but it must not be forgotten that cheese can be stored for long periods and transported easily. It provides an excellent means of converting milk surpluses into a marketable commodity, in an economic system which operated above self-sufficiency. In the course of his survey of the Lasithi plain in eastern Crete, Watrous recorded many milk-boiling sheds on the surrounding hills where shepherds manufactured cheese while their flocks grazed on the summer pastures.

The layers of soil encountered during the excavation of archaeological sites are not only sources of artefacts, bones and plant remains; they can themselves have important implications . For a number of years, a layer of'dark earth' has been recorded in the course of many urban excavations in Britain, usually in the late Roman to medieval phases. Some of these have recently been subjected to proper scientific analysis by soil scientists and botanists, which established that the 'dark earth' is an accretion of rubbish, frequently disturbed by human and animal activity, which could also be characteristic of a market garden soil. Combining the evidence of soil science with the results of archaeological excavations in London, Macphail concluded:

The possible implications are that late Roman London had a sparser population, with farming within the walls, and there had been a dramatic change from the early commercial expanding town with its entrepreneurial class. Seemingly, built-up areas had been dismantled and now were possibly used for agriculture based on dumped 'dark earth', while the legal, government and military functions of the city continued.

It has long been accepted that, by the early medieval period, Roman towns in the Rhineland consisted of little more than disjointed manors with a primarily agricultural basis, apart from royal and ecclesiastical structures. The observation that the towns of Roman Britain may have contained extensive cultivated areas well before this date has interesting implications for the understanding of the changes in the character of Roman civilisation during the first four centuries; indeed,Jashemski's work at Pompeii emphasises the fact that town and country were not as easily separable as in the modern world.

Prehistoric archaeologists have made extensive use of catchment analysis in their studies of the economies of rural sites. Essentially, the technique involves constructing a hypothetical territory of accessible land around a site, and then examining the potential of the soils which lie within it. Two major kinds of criticisms can be levelled at this approach. First, the hypothetical territories may have had little relevance to a site's actual landholdings; secondly, the present distribution of soil types may be very different to that which existed in the past, for agriculture and erosion may have had serious effects. Arnmerman has added the further point that soils which are considered unimportant in their productivity today may have been more highly regarded in the past, because of different farming practices. He also stresses the fact that catchment studies make no allowance for the inter-dependence of settlements, which could balance out their surpluses and deficiencies through exchange. Such criticisms are compounded in the Roman period, when complicated patterns of land-owning are known from documentary sources, and when trade was vigorous. Nevertheless, catchment studies do at least provide a working hypothesis which can generate interesting interpretations.

On a broader scale, soil science is a fundamental part of the geology of the world's surface deposits, particularly in geomorphology, the study of changes undergone by the landscape. Like plant remains, there is a fascinating debate about the extent to which the changes which have taken place in the last 10,000 years have resulted from climatic factors or human interference initiated through agriculture A challenging paper by Waateringe has proposed that even the soils of to forget the disastrous effects which even short temperate northern Europe could be reduced to term fluctuations in climate can cause, particu exhaustion by Roman agriculture. He emphasises that the process could take up to two centuries in the areas of the Netherlands which he has considered; if he is right, the remarkable expansion and contraction of settlement in northeastern France may be connected with this possibility. Barbarian invasions are usually blamed for the collapse of agriculture in this region.

ENVIRONMENTAL EVIDENCE AND THE ROMAN CLIMATE

Plant and animal remains tell us a great deal about Roman agriculture, whilst paterns of settlement reveal possible econminc relationships and levels of population. climate isk however, a parameter of even greater potential significance. The attitudes of historians and geographers towards climate and the limitations that it may have imposed upon human activities in thepast vary considerably, but events in Africa in the 1970's and 1980's make it impossible to forget the disastrous effects which even shortterm fluctuations in climate can cause, particularly in marginal areas. However, it is equally apparent that the origins of such disasters ~ hotly debated - how far have recent famines be caused by political economic mismanageme rather than simply by the failure of annual rain. A central issue is the extent to which climate altered by human actions such as deforestation over-cultivation, rather than independent (as therefore uncontrollable) phenomena such solar radiation or the behaviour of the earth's magnetic field.

A wide range of scientific approaches to pa climate exists, using botanical, geological as physical methods which produce a mosaic of fragmentary evidence from different parts of the world. Ideally, it should be possible to produce a agreed sequence of past changes in climate from which patterns of weather could be inferred; combined estimate of both temperature an rainfall is essential to the understanding agriculture. Using documentary evidence, supplemented by actual meteorological records for recent centuries, Lamb has worked out a convincing long-term pattern of summer and winter temperatures in England (see figure). The late medieval and modern periods in other parts of Europe also provide documentary evidence of climatic conditions which can be compared with agricultural indicators such as harvet dates yields and prices.

It is generally agreed that there has been a 'Little Ice Age' in recent centuriaes followed by comparatively rapid warming within the last 100 years; Le Roy Ladurie's pictorial evidence of the retreat of Alpine glaciers since the last century is spectacular. The previous warm period centred upon AD I1200, when the mean annual temperature in England was over one degree centigrade higher than in 1750. This apparently small difference was sufficient to extend the growing seasons of crops by several weeks during the medieval period, and in highland areas, traca of arable cultivation abound at altituda where no ploughing has taken place at any time since then. The effects on weather of thae changes in temperature are summarised very clearly by Gribbin and Lamb.

Unfortunately, the quality of documentary evidence for the climate of Roman times is patchy andg'sinadequate, although it has been used extensively (and rather too trustingly) by Lamb. Harding's summary of several different environments' climatic indicators in Europe - lake levels, pe at bogs, the extent of glaciers, etc. - is not very encouraging ; no clear pattern emerges for the early historical period. Furthemore, sources of evidence such as the rate of formation of peat bogs may reflect human interference with soil drainage rather than changing weather. It is essential that scientific conclusions about the Roman period should be derived from independent but datable sources, which are not directly influenced by human settlement or agriculture.

Temperature is one factor which has been studied by a number of different disciplines, and a world pattern has emerged which conforms remarkably well with recent historical evidence. Lamb's estimate of the temperature of England from medieval to moden times correlates well with oxygen isotope variations in New Zealand and also appears to correlate with growing-season temperatures indicated by tree specia in Michigan, although doubts have been expressed about the validity of Webb's methods. The advance and retreats of glaciers in north America and Scandinavia have been studied in great detail and dated by the radiocarbon method. The advances ofthe 'little Ice Age' centred upon AD AD 1600-1750 were of a similar magnitude to those which reached a peak around 1000BC. The intervening period of retreat lasted from c. 450 BC until the thirteenth century AD, but in north America there are signs of a short episode of glacial expansion from C.AD 700-900.

The record of glacial expansion and contraction in the Arctic has been found to be strongly correlated with information obtained from treerings. In California, the record of annual growth rings in bristle-cone pine trees extends back beyond 5000 BC, and thus comfortably includa the whole of the Roman period. It must be emphasised that tree-rings can be counted objectively, and do not rely on any other form of dating; conversely, pollen analysis, peat formation or sedimentation must be dated by the radiocarbon method or associated artefacts, which have inherent margins of error which increase with age. Treerings do not suffer from this problem; even in the Roman period they are accurate to a single year. Fluctuations in the levels of carbon isotopes in the atmosphere have been measured by Suess, and are generally thought to reflect variations in solar activity which also affected global temperature; Eddy has combined several indicators of temperaure with Suess's measurements in a single diagram (kkk)

The width of Californian tree-rings reflects the comparative vigour of their growth, and thus, by implication, the variations of climatic conditions. Their pattern correlates well with the known temperatures of historical times, as well as the much longer record of solar radiation; the peak marking the medieval warm period is particularly apparent. It is clear that the beginning of the Roman empire coincided with a peak in temperature similar to that centred upon AD ~ 200. Despite the problems of converting evidence of temperature into weather and climate, it seems reasonable to accept that the known advantages for agriculture and settlement which occurred in the medieval optimum may also have operated in the early Roman empire. Indeed, Denton and Karlen concluded that: 'Extended into the future, the Holocene pattern of climatic change implies that the Little Ice Age, if it is not already over, will be succeeded by a climate regime similar to that of the Roman Empire and Middle Ages'.

The records of the effects of the Little Ice Age are well recorded. During the peak in temperature around AD 200, the effects on agriculture in England were dramatic; vines grew in many parts of England, and fields extended well above the altitude reached by post-medieval cultivation. The population rose, and settlement extended to many marginal lands, only to be deserted in later centuries. During the Little Ice Age, the Thames froze regularly enough for fairs to be held on the ice. A study of Le Roy Ladurie's carefully collected documentary data on wine production in France has demonstrated a statistical correlation between the early date of harvests, high yields, and good wine vintage. A period of relatively favourable results recorded b these indicators occurred between AD 1500 and 1600, at which time a small peak appears clearly in the records of both tree-ring growth and solar radiation. It has sometimes been con cluded that because Roman conditions were not unlike those of today, the subject does not require much attention; this attitude obscures the vital fact that our own century has been warmer than any period since the thirteenth century AD. In fact, Roman conditions implied by the long, large peals on the graphs must have been especially favourable to agriculture, particularly m the north-western provinces of the empire.

Although it may be possible to establish a general outline of past climate, it should noe be forgotten that an outline obscures detailed fluctuations, which can only be measured accurately in recent historical periods. The winter of 1984-85 caused severe damage to olive and citrus fruit trees around the Mediterranean and in America; one such winter in the middle of a generally warm phase of the Roman period would not be detectable without lrterary evidence, but could have had severe effects upon areas like southern Spain where olive oil was produced intensively. Furthermore, temperatures have complex effects on weather and patterns of rainfall; only a few degrees difference are required to produce dramatic results.

In the introduction to his synthesis of the settlement patterns of south Etruria, Potter summarised the evidence for climate in Italy derived from the evidence of pollen samples and river valley sediments. He argued that the long continuity apparent in Mediterranean crop and stock raising minimises the possibility of gross climatic changes since later prehistoric times. Pollen evidence shows a clear correlation between forest clearance in the last centuries BC and the expansion of settlement known from finds of sites. Potter claimed that there was abundant evidence of heavy deposition of sediments in river valleys from the second century AD onwards, after a period during which it had been possible to build permanent settlements on them. He equated this sedimentation with Vita-Finzi's 'Younger Fill', an episode of silting which apparently occurred all around the Mediterranean in the late classical period . Unlike the pollen evidence, the 'Younger Fill' does not coincide with the phase of most intensive farming in central I taly which might otherwise have been held responsible for causing soil erosion through intensive arable farming. Since the sedimentation took place after the maximum extent of Italian settlement had been reached (in the late republic and early empire), Potter concluded that the erosion which it represents must have been caused by climatic change which resulted in increased rainfall.

At first sight, there would seem to be a clear correspondence between rainfall and erosion, and the reduction of human settlement in central ltaly. It might be tempting to extend this interpretation to other areas where a similar decline took place, and explain the failure of the Roman empire by means of environmental rather than political factors. There are, of course, problems: Bell has criticised Potter's extension of the 'Younger Fill' phenomenon to explain evidence for flooding on some sites in Roman Britain. Bell not only warns against extrapolating evidence from Italy to a completely different context in Britain, but also doubts that the 'Younger Fill' did in fact accumulate during a single specific period. He claims that his researches in England show that sedimentation occurred at different times in different places because of local agricultural activity rather than climatic change . Wagstaff endorses this opinion from evidence collected in Greece . This debate, and its fundamentally conflicting conclusions, highlights the dangers involved in the use of individual fragments of evidence which happen to coincide with archaeological observations in one area. and then generalising them to others.

Recent work around the Mediterranean has tended to favour human rather than climatic factors as the cause of the dramatic accumulations of sediment which took place in so many areas. By 1978, Vita-Finzi had virtually dismissed climate from the study of geomorphology. In the area of central Italy discussed by Potter, Judson has stressed the complexity of the natural processes involved in the erosion and sedimentation patterns of streams; he points out that a change towards wetter conditions might, if anything, reduce erosion by encouraging more extensive vegetation. Judson also stresses the fact that agricultural exploitation of land can increase the rate of deposition by anything from 10 to 100 times.

Giardina has shown that reports in historical documents mentioning floods and sedimentation relate directly to the pattern of forest exploitation. Thus, a recession in agriculture, marked by the reduction in the number of settlement sites in central Italy from the late Roman period onwards, led to a decrease in major floods in Rome. However, after the end of the medieval w arm period, which would have favoured the regeneration of forests, the expansion of agriculture led to a return to regular flooding, until nineteenth-century engineering works were carried out to prevent them. However, a different insight has emerged from a study of the Adige valley in northern Italy, where 'slope stability', produced by careful control of erosion by appropriate farming methods, lasted from c.300 BC_AD 600. Pope and Andel reached a similar conclusion from their work in Greece; not even an expansion of settlement in the late Roman period onto steep and marginal lands caused erosion, presumably thanks to careful terracing.

A full review of the question of erosion and climate in Greece and around the Mediterranean has recently been included in Renfrew and Wagstaff's study of the island of Melos in the Cycladc s ; there is evidence for the deposition of sediment as early as 1000 BC on Melos itself, and it occurred at varying dates and rates elsewhere. The expansion of Bronze Age population on Melos and its attendant intensification of agriculture are seen as the critical factor which removed fertile soil from large areas of sloping land and deposited it in the valley..The question of causes is left slightly open, however: 'In conclusion, the geomorphological evolution of Melos in the late Holocene can in part be related to the changes in population, economy and settlement pattern during the Bronze Age. Such a hypothesis does not preclude climatic change in later times having an effect on sedimentation rate'.

I will summarise the only general and useful conclusions about climate which are relevant to

3. The decline in settlement which occurred at various times in the western provinces of the empire (below, Chapter 5) may have been accelerated by the deterioration of soils through erosion.

4. The possibility of deteriorating climatic conditions in the first millennium AD may have exacerbated the effects of 2 and 3 above.

It would be a mistake to try to apply any single climatic model to the whole empire; as Bintliffhas pointed out, the ill-effects of a cooler, wetter phase in the west might have been advantageous in arid areas of the east. Chapter 5 shows that rural settlement did indeed continue to expand into the late Roman period in areas such as northern Syria, while it declined in Italy; is the explanation to be sought in their levels of political security, or the weather?

The Roman empire at its greatest extent contained stark geographical contrasts, from the deserts of Africa to the cold mists of northern Britain. Italy itselfexperiences great differences in temperature and rainfall between the north and the south, and its geology and landscape also vary considerably from the Alps to Sicily .It would therefore be false to expect any uniform system of agriculture to have existed in Italy, let alone in the rest of the empire.

Nevertheless, the Mediterranean climate does have sufficient uniformity to make its coastal lands capable of supporting generally similar crops. The warm wet winters and hot dry summers of the Mediterranean region have conditioned agricultural practices for thousands of years. A combination of crops including cereals, grapes and olives is grown by means of 'cry farming' techniques which still dominate Mediterranean agriculture, for which conservation of moisture is the guiding principle. Thus, whilst farmers in northern Gaul or Britain might plough their land once, to break up the surface and bury its grass and weeds before sowing a crop, a Mediterranean farmer would subj.ct it to repeated ploughings with simple wooden, in order to break the soil down finely and to keep weeds at bay, lest they remove vital moisture.

An important key to the understanding of agricul ture is the form and extent of rural settlement. ~ detailed knowledge of the nature and number a settlements of all kinds from cities to shepherds huts is vitally important if their implications fo population and marketing systems are to b' properly understood. Archaeology is of major significance in this context, but historica evidence can be more important in assessing the manner in which land was owned, managed anc exploited; productivity is not an automatic reflec tion of potential. An interpretation of the work ings of the Roman agricultural economy is very dependent upon accurate information about the size of farms or estates, the demand for their produce, and the availability of labour. A good example of the integration of archaeological and documentary evidence is provided by Peyras' reconstruction of an estate in north Africa from the owner's burial inscription, Roman law, and surviving traces on the ground . Unfortunately, Roman archaeology has tended to lag behind its prehistoric and medieval counterparts in conducting detailed analyses of settlement patterns and their environment, but auflficient advances have been achieved in recent years o allow some interesting conclusions to be drawn. The results of some of these are summarised in Chapter 5.

Each newly-conquered province had its own individual background of native agricultural traditions, which may have continued to exist with very little disturbance after their occupation . The major form of disturbance came in the form of colonies of Roman citizens, which possessed substantial tracts of farmland to provide farmsteads for their settlers in addition to the cities themselves. Colonisation was a feature of the expansion which took place during the republic rather than the empire, but it left visible traces in the form of regular land divisions, known as senturiation , which have determined the layour of fields and roads to the present day in many areas . The field survey of the territory of the colony at Cosa in Italy has indicated that the small pioneer landholdings became absorbed into larger estates quite rapidly.

Most changes in native agricultural practices probably came about through the impact of military exactions, urban markets, and the monetisation of transactions, rather than from oflicial attempts to increase output. It would be wrong to look for major technical improvements unless it could be demonstrated that output was already nearing its maximum, and that there was no spare capacity for increased production. The appearance of distinctively Roman buildings in the countryside, notably masonry villas with mosaic floors, painted wall plaster, heated rooms and bath suites, need not indicate 'Roman' owners, or a new system of landholding. There is plenty of evidence to show that the estate-owners and town councillors of Roman times were descendants of the pre-Roman aristocracy, whose status was enhanced by the adoption of the manners of upper-class Italy. However, it is apparent from the regional surveys discussed in Chapter 5 that the Roman villa estate was the principal instrument of agricultural exploitation in most parts of the Roman empire.

The most distinctive feature of rural settlement ir the northern and western provinces of the Romar empire is the phenomenon of villas. No comparable number of large masonry buildings in the countryside had ever been achieved before the late republic and early em. pire; it would not be equalled again until the postmedieval period. Unfortunately, there is extensive disagreement about the definition of villas amongst modern archaeologists and historians, which actually reflects the imprecision of the term in Roman literature : 'There is a suspicion that it is a townsman's word: that is, a villa is not simply a place in the country, but a place in the country from the point of view of someone living in a town'.

Most British archaeologists would agree about the definition of perhaps 80 per cent of supposed villa sites in Britain, and will entertain no doubts about those endowed witb fine mosaic floors and bath-houses. The problem arises over borderline cases - when does a farm become a villa? Excavation in Italy, Germany and Britain has demonstrated that many indisputable villas had humble origins, and developed gradually over several centuries from pre-Roman 'native' houses to rectangular buildings, first in timber and then in masonry or half-timbering . These modest Romanised structures might then survive as the core of a building of architectural distinction which flourished in the later empire. At what point did they become villas rather than Romanised farmhouses? A further problem is the disparity between the definitions which have been used in different areas of the empire. In the American survey of the ager cosanus, only the most elaborate buildings were classed as villas ; on Dyson's definition, many British sites would have been considered as mere farms. In Syria, Tchalenko applied the term 'villa' to buildings which undoubtedly parallel those in the western provinces, but which have no distinctly 'Roman' architectural features.

Potter's comments on the state of villa studies and the need for comprehensive excavation make it important to avoid being side-tracked into problems of definition which can obscure their real interest for the study of the economy. Taken together, their distribution must be a reflection of general patterns of landholding and agricultural exploitation, and is best studied by means of regional fieldwork programmes. Individual villas might reflect the fortunes of their owners, and the lands from which they derived profits which were at least partly spent on the construction and decoration of a villa. Unfortunately, Roman literature makes it quite clear that large landowners could own many estates, not all of which would be used as residences for any length of time. Thus, a sumptuous building might be erected upon one particular estate because the owner considered it to be a pleasant place to spend time, not because it was the most profitable in terms of agriculture. Ancient historians such as Finley have often emphasised the point that fortunes made in commercial activity were not reinvested 'productively' (i.e. in manufacturing), but were spent on land and other 'respectable' pursuits. Thus, a large villa might have even less connection with agricultural productivity, and be built from earnings made in a different activity; Gorges has suggested that the extraordinary density of villas in southern Spain was initiated by profits from metal extraction , although it is quite clear that their owners were actively engaged in intensive agriculture.

Two excavated sites will be examined here in order to explore the way in which archaeology can approach the study of villas - Sette Finestre in central Italy, and Gatcombe in south-west England. Sette Finestre has impressive architectural remains, and provides an example of the potential of an integrated programme of fieldwork and excavation, linked to historical information and broader archaeological research. It helps to give a physical dimension to historical debates about the changing patterns of land management, and the extent to which 'respectable' Roman senators could be involved in commerce. This activity was universally frowned upon in the literary sources which reflect the anti-commercial ethos of the Roman aristocracy, but was obviously practiced on a considerable scale. A complete contrast is provided by the report on a site at Gatcombe in Somerset , where a paucity rather than an abundance of evidence forced the excavator to adopt a purely archaeological approach to the site, and to frame its interpretation in the form of possibilities of different levels of credibility. The manner of its publication is almost as instructive as the site itself.

The Roman colony of Cosa and its hinterland have been studied by the American Academy in Rome . The villa at Sette Finestre has been excavated by an Anglo-Italian team, not in isolation, but together with further fieldwork up to ~984; good summaries of the villa excavation are available. The book by Carandini and Settis ~ 979 has a title (Schiavi e padroni ncll'Etruria Romana - slaves and masters/landlords in Roman Etruria which underlines the keen interest in the political implications of archaeology found amongst Italian Marxist historians. This is also characteristic of the three-volume survey of Roman socio-economic archaeology edited by Giardina and Schiavone.

The villa lies just over 3km from Cosa on a small hill in the fertile Valle d'Oro, which had originally been divided into small plots for colonists in the third century BC. The small farms of the colonists disappeared and were replaced by a number of villas by the end of the first century BC, most of which had in their turn disappeared by the third century AD . The complex of buildings on the hill at Sette Finestre has a compact rectangular plan, and includes a house (villa urbana) and farmyard (uilla rustica), with walled or colonnaded gardens and a courtyard. On the lower slopes beside the terraced access road lay a very large polygonal walled enclosure.

It is clear that Sette Finestre was not simply a recreational dwelling, for attached to the residential section is a series of processing rooms for agricultural produce, with olive and grape presses and a mill (see figure). Beneath the villa were cellars into which the grape juice could drain from the presses for wine- making; they would also provide useful cool storage for other produce. The house and processing rooms opened onto a courtyard surrounded by small rooms interpreted as accommodation for slaves. The villa rustica consisted of a farmyard surrounded by small rooms and a granary/store building. Thus, the excavated remains imply the cultivation of grapes and olives for wine and oil production, grain growing and milling, and, if the polygonal walled enclosure held animals, stock raising.

What makes the villa at Sette Finestre particularly interesting is its possible association with a Roman senatorial family, the Sestii. Lucius Sestius was a consul in 23 BC, and his name is found on brick stamps in Rome; his father Publius i known to have owned estates at Cosa in the firs century BC. The initials LS have been founc stamped on a few tiles at Sette Finestre, but the abbreviation SES occurs on wine amphorae which have been found in such quantities near the port of Cosa that it seems likely that they were made there. The same stamp occurs on amphorae found on sites and in shipwrecks in an area stretching from Cosa around the coast of Gaul and inland as far as the Rhone and Loire . If Sette Finestre itself was not the residence of the Sestii, it must represent the kind of estate-centre appropriate to a family of high status involved in commercial exploitation of intensive agriculture. The production and export of wine amphorae came to an end in the first century BC, but Lucius's brick stamps and the occurrence of SES on arretine terra sigillata of the Augustan period suggest that brick and pottery production centred elsewhere in I taly followed the demise of the wine busiress . It should not be assumed that the Sestius amphorae were actually traded by the family; amphora manufacture was not specifically related to the needs of individual estates.

This site lies less than 10 km from t centre of Bristol, beside a small river which runs parallel to the Avon into the Bristol Channel, and comprises a walled enclosure 2oom wide and at least 300m long containing many buildings. The villa itself is assumed to have lain at the south end of the enclosure, where nineteenth century writers recorded finds of building materials, and to have been buried or destroyed in the course of railway construction. Small-scale occupation of the site lasted from the later first century AD until the late third, when the wall and the buildings it enclosed were constructed . Their occupation lasted around a century, and was followed by a brief reoccupation in the late fourth century. The enclosure was divided into zones of buildings with specific functions, separated by open ground which was presumably cultivated. Finds indicated the use of one group of buildings for grain storage, milling and baking; another group was associated with the smelting and working of iron and pewter; a third group suggested storage, next to a slaughter house. Finds of domestic items and trinkets suggest that the workers may well have lived in the buildings. No buildings connected with animals were found but they may have lain in unexcavated areas, outside the enclosure.

The walled enclosure was originally thought t enclose a small town, but the location of the site the sudden appearance of the enclosure and it buildings, and the lack of any streets or planning make this unlikely. A role as a military awn factory or as the oflicial centre of an imperially owned estate is also considered unlikely from the finds. The enclosure, its ranges of buildings, and the probability that a villa building formerly existed at its southern end recalls the plans of villas known in Gaul. This layout is unique in Britain, however, and Branigan proposes that it may have been established by an immigrant Gallic landowner who had fled from the uncertainties of life in the areas of Gaul which came under barbarian attack in the third century AD - precisely the period of Gatcombe's comprehensive construction. The contrast between these tentative conclusions, and the strong literary link between Sette Finestre and the Sestii could not be more stark:

It seems that we may assert that a building with architecture of considerable quality and sophistication, fitting to be called a villa, was located at Gatcombe . . . The evidence firmly supports the construefion of the whole complex at Gatcombe, villa building included, in the late 3rd century AD . . . Neither the layout nor the scale of Main Phase Gatcombe prove that it was built by an immigrant Gaul, but they certainly make the suggestion a plausible one. (Branigan 1977, 191-2)

The concluding sections of the report on Gatcombe are concerned with the economy of the villa's estate, which is reconstructed (entirely hypothetically) according to local topography, and limits implied by the existence of other villas in the neighbourhood. Branigan arrives at a total of 6000 ha, divided between arable, pasture (upland, low Iying, and summer), sheep grazing and woodland ; the 'estate' also included at least eight 'native' settlements. On the basis of these factors, Branigan proceeds to 'discuss the way in which this land-use potential could have been integrated into a viable and profitable economy' using calculations of yields and stock capacity made by Graeme Barker and Derek Webley. Medieval analogies and modern soil science enable 1000 head of cattle and 3000 sheep to be envisaged, and a workforce of 50-60 families (with additional labour during hay-making). Wheat and barley grown in a rotation system on the arable could have supported 1000-1100 people - far more than adequate for the supposed population, and therefore leaving perhaps 76,2ookg as a surplus for export. From these entirely hypothetical figures of production from the supposed estate of the poss ible villa, Branigan proceeds to examine the archaeological evidence for the economy. He proposes a processing and organising role for the villa, through which products were channelled tc the local markets. The native settlements are seen as belonging to dependent labourers or slaves rather than tenants - 'the potential of the Gatcombe estate could only be fully realised if it was organised as a single entity'. Inward trade is represented by commodities such as pottery, 60 per cent of which came by road from more than 100km (62 miles) away from the site. Coal and iron ore were used, and Bath stone was reguired.for building; specialised foodstuffs such as salt and wine also came from beyond the estate. They were most probably acquired through town markets, where the estate's beef, grain, and other animal and plant products could be sold; official levies were probably transported to the army in Wales by way of the Avon and the Bristol Channel.

The main discussion of the Gatcombe report has been outlined in some detail because of its importance in illustrating one possible approach to fragmentary evidence, which presses the limits of inference as far as they will go - and further, in the eyes of more cautious critics. It is interesting to note a parallel between Branigan's approach to archaeology and Hopkins' to ancient history; the statement of hypotheses, and the cumulative weight of assumptions may not achieve concrete results, but can stimulate new ideas, and point out areas where new research is most urgently required. Branigan terminates his speculations with an orderly list of conclusions, ranked in order of likelihood: 'The rrixing of fact and theory, of the certain, the probable and the possible, throws the writer wide open to charges of fantasy-building and at the same time may confuse or mislead the reader who may not be clear at the end of the exercise what is fact and what is hypothesis'.

The comparatively recent revelation that precise details of plants and produce could be gained from excavation at such well known and much explored sites as Pompeii and Herculaneum is an important reminder that archaeological techniques may have much new and unsuspected information to offer. The countryside around Pompeii is famous for its large number of villas, and a part of their economic function has always been assumed to be the supply of food to the town, normally envisaged as densely occupied by houses, workshops and businesses. An extensive study of Pompeiu by Wilhelmina Jashemski has necessitated a revision of this image, for as well as the expected ornamental gardens belonging to houses, temples and public buildings, she has revealed important commercial gardens, orchards and vineyards. In exact terms, while ornamental gardens occupied 8 per cent of the excavated area, large food-producing areas occupied 9.7 per cent. As Jashemski points out:

A study of land use within the city, and of the relative density of buildings in proportion to the amount of open space, is of considerable importance in making any estimate of the size of the population. The amount of open space and the amount of land under cultivation tell us a great deal about the quality of life in this ancient city. Ancient Pompeii, with its many open areas of green gardens, parks, vineyards, orchards, and vegetable plots - must have been very beautiful indeed, and very different from the crowded, overbuilt city sometimes described by modern scholars.

The detailed understanding of the gardens of Pompeii is possible only because of the destruction of the city by the eruption of Vesuvius in AD 79, which buried it under a thick layer of hot ashes. Unless disturbed by subsequent clearance, the soil surface of that date lies immediately below the ashes; any plants, shrubs or trees growing at the time would have been covered, and were either charred by the heat or gradually decayed away. For archaeological purposes, the decay of the roots was important, for small roots left empty cavities, and larger rcot holes became filled with ashes. If the overlying ashes are carefully removed to expose the Roman ground surface, these cavities can be revealed, measured and planned. Furthermore, Jashemski demonstrated that the cavities could be used as moulds after excavation and emptying; after being filled with cement and reinforced by wires, the surrounding earth could be removed to expose the form of the root system. This allowed many species of large plants and trees to be identified, and thus assisted in determining the produce of individual gardens. Fortunately for archaeology, Vesuvius erupted in August, when many plants were bearing fruits and other crops, from which remains of seeds, beans, pips, nuts and stones survived, allowing further precision in the identification of species.

A striking example of the revision of concepts demanded byJashemski's work is provided by the open area (c.75x85 meters) immediately north of the city's amphitheatre, which had been named the forum boarium because it was thought by early excavators to have been a cattle market . Excavation and rootcavity casting revealed that it was filled by orderly rows of vines, each plant placed 4m apart and supported by a wooden stake. In addition, numerous trees were found, many between the second and third rows of vines away from the vineyard wall. Some were demonstrably olive trees, and it seems likely that the remainder would have provided a range offruits; vines could, of course, have grown up these trees as well as their supporting stakes. The vineyard also contained two outdoor dining areas with reclining benches (triclinia). The presence of many animal bones, often bearing signs of butchering, suggest. that meals were served commercially in the leaf, shade of the vines; the vineyard was located in a prime position in relation to the amphitheatre, ideally placed to attract high-class diners.

Adjacent to the owner's house in one corner of the vineyard were a wine-pressing room and a shed containing ten huge earthenwarejars (dolia) set into the ground, in which grape juice would have been fermented. It would have been possible to produce up to 12,500 litres (2750 gals) of wine. To complete the establishment, a shop front with a counter under a colonnade opened out onto the adjacent street now known as the Via dell'Abbondanza.Jashemski concludes that:

A vineyard required more intensive care than an, other Mediterranean crop but, if properly cur_i vated, it yielded large returns. The precision wit} which this vineyard had been planted and the car~ with which it had been cur_ivated suggest that the ancient vintner was a careful husbandman who enjoyed a good revenue from his vineyard which, judging from its prominent loca_tion, was a showplace much enjoyed'.

One might add that additional income from his (or her) role as rcstauratcur would have been useful, particularly as house wines and home-grown olives, nuts and soft fruits would have been available at the right seasons.

The 'cattle market' area was by no means the only commercial vineyard in Pompeii , and wine was not the only marketable product. In region I of the city, there is a building known as the 'House of the ship Europa', named after a drawing of the ship scratched onto one of its walls (fig. 8). Two houses assumed to be in single ownership front onto a street, and behind them lies an open area, 55m long and 31m wide. Approximately one third of this area nearest to the house formed a raised terrace which had been planted with young trees; larger trees were found all around the garden's enclosing wall. The lower terrace contained areas of vines, and parallel sets of beds of soil, with small tree roots along them. On the evidence of the plant remains which were found, these plots seem to have been for growing vegetables, sheltered from direct sun by nut bushes - a classic example of intercultivation.

Altogether, 240 trees were recorded, most of them young and closely spaced. Several had been planted in earthenware pots with large holes, and these have been interpreted as an indication of grafted species, probably soft fruits, cultivated in a manner described by Pliny the Elder.

Other specialised gardens studied by Jashemski at Pompeii include a large orchard, and an establishment where flowers may have been grown in well-watered beds shaded by large trees, probably for the production of perfumes as well as the flowers themselves. The obvious implication of so much commercial rather than purely ornamental gardening, in the final years of Pompeii at least, is that the kind of agricultural production hitherto assumed to have taken place on farms and villas near to towns could in fact exist within them as well. The distinction between town and country is therefore truly blurred.