The Creative Use of Bias in Field Survey
Bradley, Richard, Durden, Tess, Spencer, Nigel, Antiquity
A modest experiment explores what is seen and what is not seen in field survey, and what can be done about it.
The Editor of ANTIQUITY recently described a sensation that must be familiar to many archaeologists. It is 'the special moment . . . when you feel at last you have grasped the pattern, the logic of the landscape and the logic of the archaeology within that landscape' (Chippindale 1991: 442). He cited the way in which prehistoric rock-carvings in northeast England are almost always discovered in the same kind of locations -- on the edge of low hills where the stone is exposed along the break of slope. Once that is appreciated, the carved rocks are easy to find. His account certainly captures the excitement of discovery, but it could also illustrate the ease with which we may become the prisoners of our own expectations.
If the excitement of discovery seems familiar, that danger must be familiar as well. Such preconceptions are a particular feature of extensive field survey where they assume respectability under the guise of 'judgement sampling'. These biasses have been recognized for a long time. We know that individual excavators may find different kinds and quantities of artefacts (Clarke 1978) and that field surveyors too may recover different classes of material. Shennan (1985: 40-44), for example, has shown how the individual members of his team tended to find either lithics or pottery, but did not recover both kinds of material in their true proportions. Field-walkers have also been known to recover pottery of one period at the expense of artefacts belonging to other phases (Guy Sanders pers. comm.). Similar biasses may exist on an even larger scale. In Mediterranean archaeology Cherry has claimed that different kinds of survey may encounter quite different kinds of site (compare, for instance, Cherry 1983 with McDonald & Hope-Simpson 1961). In the New World there have even been claims that certain survey methods may help to create the very 'sites' that they study (Plog et al. 1978: 414-15; O'Brien & Lewarch 1992: 271-5).
In each case a major problem lies in our ability to impose order on our observations -- an individual field-worker may become attuned to finding a particular class of material. It is all too easy for the archaeologist to recognize one kind of patterning at the expense of others, for that seems an inescapable feature of human perception. Layton illustrates this problem most effectively when he compares different records of Palaeolithic art (1991: 26-33).
One way of redressing these biasses is by increasing the intensity of surface cover and implementing an explicit sampling design (Cherry 1983) Another approach is to monitor the changing quality of the observations made in the field, recording the results produced by individual workers, or the light conditions under which survey is conducted (Hodder & Malone 1984: 123-32). In principle this can allow any systematic errors to be recognized and measured, but it also takes us far from the ideas that structured the survey in the first place. As in excavation, documentation can easily become an end in itself until the aims of the project are obscured by its methodology.
Using biased samples
If such biasses are difficult to eradicate without losing sight of our wider objectives, it could be more efficient to put them to use. What if the same areas are searched by several teams, each responsible for collecting different categories of material. By focussing their attention on single classes of data, it should be possible to improve recovery rates and to offset the problems raised by selective recognition. Of course much depends on the scale of the survey. The smaller the project, the harder it will be to eliminate the personal element, but in the case of extensive survey this bias might be employed to search the landscape according to alternative hypotheses.
In many cases the main objective of rapid surveys is to uncover the logic behind the location of sites, so it would be valuable to exploit the biasses that seem inevitable. …