In 1997, we dug eight test pits in an attempt to clarify what the geophysical surveys seemed to be telling us. As well as helping to determine the total depth of deposits across the site (and how close to the surface the bedrock lay), they would also give us some idea about how the deposits had formed and what they consisted of. Each pit was only one metre square and each was dug by hand down to bedrock (or, where the base of the pit became choked with boulders, down to a point where it was no longer possible to excavate).
The test pits showed a similar picture over most of the site: the bedrock had weathered into a crumbly orange sand, often with large, irregular lumps of sandstone débris. Over this was a pinkish layer containing Mesolithic material and less of the sandstone rubble. This was then sealed by a topsoil containing mixed prehistoric and post-medieval finds over which the grass was growing. In 1998, we were able to refine this slightly. We extended Trench I to the west, down the slope, in a strip one metre wide. The topsoil was found to consist of two distinct deposits. The uppermost contained a great deal of root material and very little Mesolithic flint, while the lower appeared slightly less grey, had fewer roots and more prehistoric material.
As a result of the test pits, we now have a much clearer idea of the formation of the site. The initial erosion of the cliff face took place in the immediately post-glacial period, probably before the first Mesolithic occupation of the shelters. A soil then formed on top of this, which was probably the late Mesolithic topsoil. Erosion of the cliff face was much less apparent and perhaps more sporadic at this time. Finally, in the six thousand or more years since the late Mesolithic activity, there has been little new soil formation and most of what there has been probably occurred as a result of the gradual washing of deposits down the slope.