modes of dispersal quite different from those current at
any one point in time. The full extent of a pottery type’s distribution
irrespective of differential periods of use in certain areas is
nevertheless of considerable interest; the maps here presented should be
studied with reference to the main text in order to ascertain whether such
differentiation occurred.
The present author’s maps seek to relate the distribution
of pottery types to natural and man-made features of the Romano-British
landscape (see also Pollard 1983a, 415—73). The non-random behaviour
that such features may induce need not always be apparent in the spatial
patterning of artefacts or sites, however (Hodder and Orton 1976, 9).
Cumulatively, the range of factors that could have influenced the
dispersal of pottery may produce a random pattern. A subjective approach
to interpretation might result in the false identification of structure in
random patterns (ibid., 4—8). The objective search for non-random
patterning of artefact and site distributions has been one of the major developments
in spatial analysis in recent years, but it should not be overlooked that
the identification of such patterning is not an end in itself. The
processes that resulted in random or non-random patterns remain to be
identified; ‘one spatial pattern may be produced by a variety of
different spatial processes. . . .often one must look to non-spatial
evidence to corroborate or disprove theories about spatial processes’ (ibid.,
8).
The distribution maps published here have not been subjected
to tests for randomness. There are two reasons for this omission. Firstly,
it was considered (subjectively) that the patterns could be interpreted
with reference to kiln/production area location in many instances (e.g.
Fig. 22, 48) or to the operation of other factors (Pollard 1983a, 415—73)
which have been investigated by previous research (e.g. Hodder 1974b;
Loughlin 1977). Secondly, the nature of the area under study, with the
coastline a dominant feature of the geography of Kent, places severe
limitations on the value of nearest-neighbour analysis in
|
|
searching for randomness (Hodder and Orton 1976, 41—3). The approach
adopted by Hodder and Orton (1976, 44—5) to overcome this problem
would be inappropriate to the present study, as it involves the choice of
area for analysis such that the nearest-neighbour distance from any site is
less than or equal to the distance from that site to the coast. This would
seem to rule out many of the most important sites studied by the present
author, such as the military bases at Richborough, Dover, Port Lympne and
Pevensey, and the civil settlements at Little Shelford (Essex) and
Birchington.
It was initially hoped that the comparative efficiency of the
major production units in Kent and the surrounding area in supplying their
potential hinterland zones might be studied through quantified statistics.
Work by Hodder (1974d) suggested that regression analysis would be an
appropriate technique: at a generalised level the fall-off gradients thus
generated may reflect the comparative ‘value’ of pottery types, the
shallower gradients being equated with types which were marketed in larger
quantities at greater distances from their production centres. However, this
analysis was curtailed by the weakness of the quantified data base. An
alternative approach to regression analysis, involving the measurement of
the density of sites with the pottery type under review in bands around the
type’s source, would seem to be inappropriate in the case of Kent owing to
its maritime location; the density of sites greater than 10 km. from either
Canterbury or the Cliffe Peninsula area kilns (producing BB2 and grey wares)
would inevitably show a marked fall-off due to the inclusion of the Thames
Estuary and English Channel in the bands. In the event, sufficient
quantified data, expressed in the form of the proportion of one pottery type
in assemblages, was collected for one small study to be undertaken (Fig.
48). The number of points was considered to be too small to warrant detailed
analysis, but a marked fall-off in quantities outside 10 km. radii from the
BB2 kilns around the Cliffe Peninsula is
|