FIELD WALKING
FIELD WALKING METHODOLOGY
THE PROJECT
The fieldwork for the project began in the two weeks immediately before Christmas, 2004, with set-up and training by Trent & Peak Archaeological Unit (now Trent & Peak Archaeology, TPA) of the University of Nottingham, who had been contracted as advisors to the project. The development and implementation of strategy and day-to-day running of the project was the responsibility of BHTA. On resuming work after a break for Christmas the project was run entirely by the volunteer field walkers and this continued to the end, requiring only minimal intervention by TPA. Specialists from TPA and independent consultants were used to identify the finds.
Though initially planned as a three-year project it became evident early on that the rate of progress was too slow for the project to be completed in three years without changes to the field walking protocol. After two applications for extensions were granted the project was completed in five years. In all it took 74 weeks of field walking to complete the project at an average rate of survey of between 11 and 12 hectares of arable land a week. 868 hectares were field walked.
Although the project was initially limited to the parish, at the end of fieldwork it was realised that important information needed to interpret a field at the edge of the parish was available in a field beyond the parish boundary. This field (number 2890), on the eastern side of Granby Lane on the far east of the parish and straddling the Aslockton and Whatton parish boundary, was walked during two weeks in November 2008, but only the Iron Age, Roman and Anglo-Saxon finds were processed.
The objective of the project was to carry out a systematic field walking survey of the whole of the parish to a common standard with the aim of providing a detailed history of settlement and land use. In detail the objectives were:
- To carry out a reconnaissance by field walking to identify sites and areas for potential future detailed investigation
- To carry out a full analysis of finds types, dates and locations and distributions to enable the preparation of phase maps showing settlements or other areas of activity by period.
- The detailed aims can be summarized as follows:
- To locate evidence of prehistoric occupation of Bingham.
- To establish the full extent of Roman activity in the parish
- To find evidence of Saxon settlement
- To throw some light on land use and settlement patterns in the medieval to modern period.
METHODOLOGY FOR FIELD WALKING
When planning for the grant application it was decided that that the project could only be completed in the three years that the grant covered if a 10% survey was carried out. However, during the training period we were encouraged to adopt a 20% survey; i.e. 2-metre wide transects 10 metres apart with 5 metre stints, which is common practice elsewhere. However, it became evident during the first field-walking season that the project could not be completed in the time for which the funding was allocated at this rate of ground coverage. The survey was changed to a 10% survey with 2m-wide transects 20 metres apart during the first season. During the second season it was realised that even with the changed protocol the pattern of crop rotation was such that there was a finite number of fields available each year and that the project could still not be completed in three years. An application to the LHI to extend the project to four years was accepted. A further application to extend the project to five years in order to complete the interpretation was also granted.
The protocol for field walking adopted by the project is described in full elsewhere. In summary, field walking took place during the winter when the crops were dormant. In each field a base line was laid out at right angles to the direction in which the transects were going to be walked to enable maximum coverage of the field. The transects were 2 metres wide and because this was likely to be a once-only parish-wide survey it was decided that field walkers should pick up everything that was man-made or otherwise significant, regardless of age. Each transect was divided into stints 5 metres long and all finds from each stint were put in one bag. The transect letter, stint number and collector’s initials were marked on each bag in the field.
Field walking was done in three-hour sessions, two a day Monday to Friday and one on Saturday. Field walkers from a pool of about 45 volunteers were canvassed every Thursday to state their availability for the forthcoming week and enable a rota for the week to be set up.
After collection finds were processed as follows:
- Bags were emptied and the finds set out to dry in fruit and vegetable boxes (collected from Bingham market)
- Finds were washed and dried
- Each find was put into an individual bag on which the stint number, transect letter and collector’s initials were transferred
- The relevant data were entered into the database. Each find was given a unique number allocated in the Access database. This was marked in red on the bag.
- Finds are individually numbered
- Each find spot is given a 12-figure grid reference that is entered in the database
- Finds were divided into the various categories used in the project and some sent away to be identified.
All the information on finds was entered in a relational database in Access. The mapping software used was ArcMap, a subset of ArcGIS 9. Some of the processing of data was done in Excel.
The main topographical reference is the modern O.S. 1:10,000 scale digital map. The modern fields have been given unique numbers that are derived from the national grid reference for a point within the field. To aid the analysis of the data collected reference has been made to four earlier maps, all of which have been available in digital form and have been warped to fit the National Grid. These are a conjectural map of the parish for 1586 that has been constructed in this project, an estate terrier produced in 1776 and also compiled in this project, the tithe map of 1842 and the 1883 OS 6-inch map.
CROP CONDITIONS
The main crops grown during the period 2004 to 2008 were wheat, oil seed rape, barley and sugar beet. The crop rotations used by the farmers vary. On Holme and Starnhill farms there was a two-year rotation between oil seed rape and wheat. The other four farms employ longer rotations, with occasional interventions of crops such as field beans, potatoes, linseed and peas. Oil seed rape provides a particular difficulty for field walking because it quickly covers the ground, leaving very little time when the soil remains exposed. For three seasons the whole of two of the farms, about 1000 acres in total, were given over to oil seed rape, one farm in 2004/05 and 2006/07, the other in 2005/06.
Two farms, Starnhill and Holme do not use a plough. Instead, they use a cultivator, which works only the top few centimetres of soil, leaving the stubble partly buried. Once the ground is drilled the soil may remain partly covered by stubble and chaff, which hampers visibility.
A substantial portion of Holme Farm is rich black soil in which growth is quick. In the mild autumn of 2005 the whole area of wheat became so thick that field walking had to stop on19th November, though it was possible to continue on other farms. A less serious problem was posed on Starnhill Farm, where field walking was possible only until mid January in 2005, after which the crops were too thick.
RELIABILITY OF RESULTS
There are five main factors that have an influence on the reliability of the results obtained from field walking:
- Inexperienced field walkers
- Collector bias
- Ground visibility
- Soil type
- Methods of cultivation
None of the field walkers had any experience of field walking before the project started and though many of those who began at the start continued to the end of the project, there was a steady flow of new volunteers throughout the project.
It was noticed that some field walkers saw less than others working in the same field and that this did not change significantly with experience. It is believed that collector bias and a regular turnover of volunteers were both compensated for by setting teams of several individuals to work in each field, always randomly allocated to transects.
Most of the fields walked were winter-sown wheat or barley. Normally, the crop becomes dormant during cold spells, but in the winter of 2005/2006 the weather remained warm enough to allow growth intermittently throughout the winter. As a result, ground visibility was significantly reduced towards the end of the season and some fields were walked at the limits of viability, which was about 50% visibility. Some of the wheat and sugar beet was spring sown.
In most of the area walked the soil is a sandy loam that weathers well, but areas of clay in the south-west and north of the parish do not break down as well, probably concealing artefacts in clods. These factors cannot be compensated for and comparisons of, say, average finds per hectare, between fields with different soil types and walked at different times during the growth cycle must be treated with an element of caution.
STATISTICAL ANALYSIS
Standardisation of collected data
At the time that the decision was taken to change from a 20% to a 10% survey an analysis of the finds from field 2388 showed that none of the pottery and flint types collected during the 10-metre survey would be lost by removing alternate lines to simulate a 20-metre survey, though the relative abundances were different. As a result the data collected in the fields done at 20% were normalised to 10% by removing alternate transects from the database and processing the fields using only the residual data.
Finds densities
The standard procedure followed by archaeologists in many parts of the country is routinely to do 20% surveys and quote finds densities on a field-by-field basis as finds per hectare for a 10-metre survey. It was decided to test the efficacy of making comparisons between Bingham and other areas by doubling the figures acquired on a 10% survey. The two fields chosen for the test are 0007 and 1786, both of which were surveyed at 10 metres. Thus, in the Tables1.1 to 1.3 the two columns show the finds that were collected on two sets of alternate transects: one set has been retained in the database; the other set has been excluded.
In all three cases fewer fabric types were recorded in the removed transects, which means that in these two fields the full range of fabric types collected in the 20% survey remained in the database. This may have been accidental, because in all three tables there was not a perfect match across the columns. If the finds had been evenly distributed in a field the totals for the two columns should be the same and there should be equal numbers of each of the fabric types in each column. In Table 1.1, where the total for each fabric type is low, there is a reasonably good comparison. In the other two tables there is little equivalence and the totals are significantly different except for coarse earthenware in Table 1.2 where there is a variance of less than 6%.
Table 1.1 Medieval fabrics in field 0007
20m spaced transects left in database |
20m spaced transects removed from the 10 metre-spaced survey |
|
| Fabric type | Number of finds |
Number of finds |
| Medieval glazed | 2 |
2 |
| Medieval sandy | 1 |
1 |
| Nottingham green glazed ware | 1 |
1 |
| Gritty ware | 3 |
0 |
| Light-bodied gritty ware | 1 |
2 |
| Midland purple | 2 |
5 |
| Total | 10 |
11 |
Table 1.2 Post-medieval fabrics in field 0007
20m spaced transects left in the database |
20m spaced transects removed form the 10 metre-spaced survey |
|
| Fabric type | Number of finds |
Number of finds |
| Black slipware | 3 |
2 |
| Black ware | 2 |
9 |
| Cistercian ware | 6 |
0 |
| Coarse black ware | 8 |
12 |
| Light-bodied black ware | 1 |
3 |
| Mottled ware | 9 |
13 |
| Purpled-bodied coarse earthenware | 1 |
0 |
| Sandy coarse earthenware | 1 |
1 |
| Slipware | 6 |
11 |
| Yellow coarse ware | 3 |
6 |
| Yellow ware | 7 |
6 |
Total |
47 |
63 |
| Coarse earthenware | 137 |
129 |
| Total | 184 |
192 |
Table 1.3 Medieval fabrics in field 1786
20m spaced transects left in the database |
20m spaced transects removed form the 10 metre-spaced survey |
|
| Fabric type | Number of finds |
Number of finds |
| Early Medieval fabrics | 7 |
0 |
| Medieval glazed | 1 |
0 |
| Medieval sandy | 6 |
6 |
| Nottingham coarse pink/orange sandy ware | 9 |
5 |
| Nottingham green glazed ware | 12 |
4 |
| Nottingham light-bodied green glazed ware | 4 |
2 |
| Gritty ware | 5 |
3 |
| Light-bodied gritty ware | 12 |
4 |
| Midland purple | 10 |
6 |
| Total | 66 |
30 |
Calculating the finds per hectare using these different approaches produces the results shown in Table 1.4.
Table 1.4 Comparison of finds per hectare for sets of alternate transects in a 20% survey
1 |
2 |
3 |
4 |
5 |
| Field number | Date range or fabric type | Finds per hectare 10% survey recorded in the database (20m) | Finds per hectare 10% survey excluded alternate transects (20m) | % variance between columns 3 and 4 |
| 0007 | Post Medieval | 9.87 |
13.23 |
+34 |
| 0007 | Coarse earthenware | 28.78 |
27.1 |
-6 |
| 1786 | Medieval | 3.91 |
1.78 |
-54 |
The variation between the finds densities calculated from two sets of alternate lines is least for coarse earthenware, which has the highest densities, but for the other post medieval wares (+34%) and medieval wares (-54%) the variance is large. Doubling either of these sets of figures to simulate a 20% survey would be acceptable for the coarse earthenware, which is abundant and evenly spread over the fields, but for the other two types would be misleading. For example doubling the density for transects that are included in the database (column 3) for post-medieval wares would give 19.34 fph compared with 23fph for the full 20% survey. This is a variance of 16%. For the medieval wares the variance is 37%.
It must be concluded that where the unit for analysis is a modern field comparisons within the project area and with areas outside it can only reliably be made if the survey protocols are the same; i.e. compare only 10% surveys with 10% surveys etc.
Except where there are very high densities of evenly spread material reliable results cannot be obtained by doubling the figure for a 10% survey to simulate a 20% survey or by grossing up to simulate a 100% survey.
Sample areas
It is common practice among archaeologists to quote densities of finds as finds per hectare for each field surveyed. Modern field boundaries often bear no relation to the field system that was implemented at enclosure. They have no relevance at all in prehistoric times or during the period of open field farming, except where there may have been pre-existing closes.
In this study for periods before parish enclosure, which took place approximately 1680-90AD, the finds densities are shown with respect to a 100m square grid fitted over the whole of the parish coincident to the National Grid. Each square is one hectare. This method provides a common, uniformly sized, independent sample unit for the whole parish and, subject to the variables discussed elsewhere, the method enables comparisons to be made between different parts of the project area.
For the post-enclosure period, finds densities are given with respect to the field system that was appropriate for the time. The basic unit chosen is either the field shown on the 1776 map for studies of 18th century finds distributions or the field on the six-inch to one mile OS 1883 map for the 19th century. There are some differences in field boundaries between the 1776 map, the1842 tithe map and the 1883 map. These are taken into account where necessary. However, it is assumed that with only minor differences the 1776 map reflects the original enclosure of 1680-90.
For representation on maps the values of the finds per hectare are classified using a facility within ArcMap. It provides a choice of six methods, but the default method is defined as:
Natural Breaks (Jenks): Classes are based on natural groupings of data values. In this method, data values are arranged in order. The class breaks are determined statistically by finding adjacent feature pairs, between which there is a relatively large difference in data value.
Throughout this study the natural breaks (Jenks) method has been used. The class order is represented by the same colour hierarchy on all maps.