Evaluation of Archaeological Decision-making Processes and Sampling Strategies

This pilot study, undertaken by the Oxford Archaeological Unit at the behest of Kent County
Council, has examined the archaeological decision-making processes and the actual and
potential sampling strategies on some major infrastructure projects carried out in southeast
England in the last decade. The project was funded by English Heritage and the
European Regional Development Fund, as part of its Interreg IIC programme.
The twelve projects selected for study provided a range of types and periods of archaeology
in a variety of topographical circumstances with diverse land-use histories. The c 240
hectares covered by these projects had been evaluated by a suite of techniques, and very
large areas had subsequently been examined and planned during excavation and watching
brief. Thus it was possible to compare the predictions made at evaluation stage with the
remains encountered in fieldwork over a very large area. A crucial element of the study was
the computer simulation of alternative trenching strategies upon digitised site plans,
including different types of array and sample sizes.
In addition to the OAU analyses, Archaeometry Branch of English Heritage’s Centre for
Archaeology undertook a more detailed study of the geophysical surveys undertaken on five
of these sites.
All non-intrusive methods of evaluation had merits in certain circumstances, for example
desk-based assessment for developing effective strategies for evaluating sites, fieldwalking for
locating sites with durable artefactual remains and prehistoric sites that only survive in the
ploughsoil, and geophysics for revealing remarkable detail about feature layout for those sites
with magnetically-enhanced soils. These methods were all comparatively cheap, but they all
had some serious failings and none were even moderately successful at evaluating the range
of archaeological remains that survived on these projects. Machine trenching was the only
effective means of predicting the character of the sites in this study and, even though it was
more expensive than other methods, the improved quality of information and greater
certainty from which to devise a mitigation strategy, made it cost effective. In practice, all
the projects adopted more than one technique of evaluation and the combination of
judiciously selected methods proved to be a powerful predictive tool.
Eleven of the projects within this study had been evaluated by machine trenching, at samples
of between 0.8% and 5.6%, the average being 2.4%. The simulations suggested that the
proportion of the sites seen in evaluation was too small to predict with confidence the full
range of archaeological material actually present upon them, and this conclusion is borne out
by the unexpected discoveries made on the sites when they were stripped to examine
remains of other periods. The percentage of a site that needs to be seen to assess
adequately the extent and survival of archaeological remains depends on the character of
the site. Where linear boundaries, substantial features and clustered remains survive, and
Roman sites are obvious examples, a lower sample could be adequate, though even here 3%
- 5% would be required to expect a moderately good assessment. However, more scattered
and ephemeral remains, and Bronze Age and early medieval settlement sites are good
examples of these, could be missed entirely by sampling at this level.
Chance evidently plays a part in site detection when trenches are placed without any
knowledge of features below the ground. Experiments to assess the range of variability that
can arise by the systematic but random location of trenches suggest that sampling at a given
fraction can reveal up to 1.5% more or 1.5% less of the archaeology on each site than the
expected proportion. This clearly makes sampling at 2% a high-risk strategy.
Trenches on the projects in this study were laid out in a grid formation, or in regard to
particular features seen on air photographs or in geophysical survey, with the exception of
one site that had a customised design. The grid pattern with single-width trenches 30 m or
20 m long proved to be the most effective design, along with parallel trenching, although
when sample proportions reached 10% there were fewer differences between the different
arrays. The size of the gaps between trenches was the most important element in trench
design.
This study indicates that the single most important factor in the success of evaluating
archaeological sites is the date of the remains that survive upon them, and this is true
regardless of the character of the geology and topography, depth of overburden and recent
land use, and it is true for all techniques of evaluation. The methods we commonly use are
successfully locating Roman, medieval and, to a lesser extent, Iron Age remains, reinforcing a
known bias in the archaeological record, but those of Neolithic, Bronze Age and early
medieval (Anglo-Saxon) date, landscape features and those on topographies where
settlement was previously thought to be absent are only being revealed as a result of
extensive stripping in large infrastructure and construction projects. This suggests that we
are consistently missing sites of this character. The benefits of large-scale stripping were
apparent within the projects that formed part of this study, and this work suggests that
serious consideration should be given in the right circumstances to stripping, planning and
sampling sites (strip, map and sample), with further follow-up work concentrating on
critically selected areas.
The pilot study was able to examine only twelve projects which had very diverse
characteristics, and hence the general trends that can be seen in the data cannot be validated
statistically. Nevertheless, it raises important issues worthy of further investigation. It is
hoped that this work will prompt further studies of this kind.