A lot of archaeological work has been done on Roman military sites, both in Britain and across the territory of the former Roman Republic and Empire. This work began with the dawn of archaeology as an academic discipline (indeed, it began before the development of, for example, the concept of stratigraphy) and has often been instrumental in developing the archaeological discipline through the years. Much is known about how Roman military sites, particularly forts (sites for smaller units of the Army such as auxilia, numeri, etc) and fortresses (sites built for the larger units of the Army, the legions) developed, progressed, changed, and were eventually abandoned. The frontier systems of the Empire, in Britain and beyond, are well studied. We know the names of most of the buildings in the forts or fortresses, and implicitly from those names, their function. There is an amazing amount of finds from these sites, and after the advent of academic archaeology, most of those finds have been dutifully recorded and cataloged after excavation.

Archaeological excavations and subsequent recording, classifying and cataloging of excavated artefacts can give good insight into life in a Roman military installation and its occupants' use of those artefacts. But that insight can be refined and expanded. One way to do that is to use the power of modern computers and aggregate data to examine aspects of the artefacts, their functions, and their findspots. This examination can give us a more detailed idea of the daily lives of Roman soldiers, and that is what this study has attempted to do.

Theory and Broad Goals

The theoretical background of this study arises from Behavioral Archaeology, which itself arose from the New and Processual Archaeology paradigms created in the 1960s and 1970s. Michael B. Schiffer, the original proponent of Behavioral Archaeology theory, suggested that archaeological artefacts are variable in four dimensions, all of which are part of the archaeological record's formation processes. These four dimensions are the formal, the spatial, the frequency and the relational.

The formal dimension of the artefact is its shape, size, type, color, weight, material. All of these are generally well studied with respect to Roman artefacts. Functional groups, used in this study to divide artefacts into groups based on their functions which give indications of the activities prevalent in the areas in which they were found and the essentially, the overall functions of those areas, are part of this dimension. For more on these groups, see the functional group page on this website. The spatial dimension of the artefact is its location in the archaeological record of the site to which it belongs. In this study the spatial dimension was further analyzed by assigning the artefacts' find locations to sectors in the military site/forts. For more on this, see the sectors page on this website. The frequency dimension is the number of times the artefact's type occurs in the archaeological record. This dimension was also manipulated in this study, giving rise to percentages and percentage means for each functional group. And the last dimension, the relational dimension, consists of patterns of co-occurrence within the artefacts, also called associations. The relational dimensions of artefacts, and the systemic behavioral contexts which give rise to them is something that is examined in depth in Roman Soldiers and the Roman Army, and in the composite fort presented on this website.

Using computers to examine archaeological data, while rather commonplace now, was something very new in the late 1970's and for another 20 or more years after. In 1977 Stanley South published his work on sites in colonial America (republished in 2002). He used a main frame computer to expand his understanding of the sites he had excavated. South used some of the theories behind Behavioral Archaeology to further elucidate his work, while developing a procedure which basically saw the artefacts from the sites divided into formal dimension functional groups and assigned a spatial dimension locational variable depending upon where they were found. Because these sites were very similar in layout it was possible to use repetitive location variables across each site (for instance, kitchen, living quarters, outside kitchen door, entryway, etc) to gain insight into how that location was used by the original inhabitants of the site.

This method of analysis, developed by South and backed by the theories of Behavioral Archaeology, was not often, if ever, replicated after South's study. Perhaps this is because as I, the author/webmaster of this site, have discovered, entering and then analyzing large amounts of archaeological data into either a main frame or even a desktop computer is time consuming and can be very tedious and remains so, even in the modern computing era. Analyzing the data so that the finds are put into functional groups, and their findspots are put into sectors (of a military site, or a villa, or a village, or a farmhouse, etc) is also very time consuming, and must be done anew for each archaeological report. Having said that, I decided to follow and continue to follow the method developed by South, to modify and enhance it as needed for my purposes, and use it to examine finds from Romano-British military sites. My specific aim was to examine the spread of finds across a 'typical' Roman military site to understand how items found archaeologically from each location within the site related to the purported purposes of the buildings or areas in that location. I used an idealized Roman 'fort', the composite fort, to do this examination.

Britain has many Romano-British military sites, which range over a large part of the island, and this study concentrates, for now, on the military sites in the northern part of the British province; those from northern England and southern to mid Scotland. Quite a few Romano-British military sites (but by no means all, or even most) have been excavated in the last 150 or more years. Much of what is known about these sites, and the Roman occupation of Britain, comes from excavations which were done well before computers; before the time of electronic recording and publishing. The records of those excavations are in danger of being disregarded in new studies, because they are not available electronically and much new research and publication is being done, at least partly, electronically. Therefore, another goal of this study was to take published archaeological reports and move the data from those reports into electronic form and to see if the data from older excavation reports, perhaps even the oldest, can be made useful in a computerized format.

This goal was more difficult to accomplish than originally envisioned. This is mostly because the earliest archaeological reports are often unuseable. If the excavation took place before, for example, the elucidation of samian forms, then the pottery report from that publication is more or less useless for this study. The understanding and usage of basic archaeological premises, such as stratigraphy, in some of these early reports is limited or non-existent. However, if a site was dug in this early period, pre-1920/30's, but was then reanalyzed in a more recent period (such as Bar Hill), the results can be quite useful and have been included here.

Another factor which has made using older reports a bit difficult is the changing trends in archaeological reporting of artefacts. For many years, for example, animal finds were not reported in depth. Other trends include limited reporting of ceramics, such as amphorae, or changes in how ceramics are counted (rim sherds versus all sherds, etc). And, of course, 'uninteresting' finds, such as native ceramics, are not always reported.


A bibliography of published archaeological reports was created for each site examined. Non-published reports are welcome to be included in this study now, but were not used in the original study due to access and time constraints. Each report for a site was read and all the artefacts discussed therein were entered into a computerized database. This database is created with mysql and php and aspects of it are available on this website. Originally mysql and spreadsheets were used for data entry.

To enter the data a thorough understanding of the excavation had to be gained. The artefacts were assigned periods (occupation periods) of the site according to their findspot levels if that information exists. The artefacts were also assigned to a sector of the fort, again depending upon their findspot. And depending upon their type, the artefacts were assigned to a functional group and a sub-group of that group. For more on the military site, or fort, sectors see the sectors page. Functional groups are discussed on the functional groups page.

Stratigraphic context was divided into three categories for each functional group, depending upon the completeness of the published artefact catalogs about the site. These categories are 'All' which includes all the functional group's items from the sector; 'Strat' which includes only those items in the group which were found stratified; and 'Good' which includes stratified items in the group from excavation reports with artefact catalogs which were considered complete (i.e. every artefact of that type from the site was included in the catalog instead of only a representative sample).

Using the database functional group percentages were computed for each site and that sites's sectors. These percentages are those of each functional group compared to the overall number of artefacts from each fort sector (You can see the each site's particulars on the site percentages page). Using these percentages from each site, the percentage mean (average) of all the sites considered in the study for each functional group and for each sector was computed. These were used to form the composite fort, discussed in the composite fort section of the website.

The composite fort was developed in an effort to compensate for the often limited extent of the excavations of individual forts (with respect to the sizes of those forts). Hopefully all the finds, separated into groups according to their functions with their percentage of total finds computed into a percentage mean for each functional group can give a picture of the 'average' life of the soldiers and their use of the objects which became archaeological finds. Also, comparing an individual fort to the composite fort can give insight into the ways that fort may differ from an 'average' fort, providing clues to activities which may have been specific to that fort.

The 'AP' occupation period highlighted on the sites and composite fort webpages refers to a constructed artificial period in which all the finds and their functional group percentages from a site or percentage means from the composite fort are featured, no matter what the occupation period in which the artefacts occurred. The AP composite fort gives a idea of the overall patterns of artefact functional groups in Roman military sites, which is useful. But it does not show how artefact functional group patterns may change over time. To examine those changes you may also explore the composite fort by individual occupation period. That ability is coming soon for the individual site results.

It should be noted that in the original book, Roman Soldiers and the Roman Army, and throughout this website the Roman military sites discussed are referred to as 'forts'. This is because most of them are forts, as the term is used in Roman military studies. However, vexillation forts, fortresses, fortlets, milecastles and other military sites can be analyzed in exactly the same manner as used here, with exactly the same parameters. This is because these sites have the same types of sectors in either greater, or much lesser numbers than the 'fort', and the same types of artefacts are found in them. Eventually it is hoped that data from these other types of military sites will be added to the database in order to examine how, and if, they differ from the 'fort' site.