suggested. It would be interesting to examine whether this pattern is
confirmed by the excavation and analysis of suitable groups of pottery from
other sites of the same period and from sites with somewhat earlier
assemblages. The occurrence of large quantities of poor quality imitation
BB2 from Greenhithe (Detsicas 1966; nos. 107—109 here; 4.111.2) in a
Hadrianic-early Antonine context suggests that the pattern might be somewhat
different in this period. The inclusion of Essex material would also be of
interest.
The relationship of distance to dispersal of kiln-products has
also been investigated by plotting the number of forms of a certain fabric
occurring on sites in the study area (Figs. 10 and 11). Empirically, it
would appear that the ‘value’ of a pot varied according to its
form/function (Pollard 1983a, 415—73) with the result that higher-value
types were distributed over a greater distance than lower-value types.
Variation in the number of types present on sites could thus be interpreted
as reflecting variation in the economic distance from the source. This may
reflect the prosperity, topographical situation, or distance in miles from
that source. In the case of Brockley Hill ware, a marked fall-off east and
south-east of Southwark is apparent, corresponding to an increase in mileage
from the source (north-west of Southwark). In Kent, it is possible to
distinguish concentric zones wherein certain numbers of types are to be
found irrespective of site-type: 4—6 types west of the Cray, 2—3 types
between the Cray and the Medway, and 1 type only (usually mortaria) east of
the Medway. This zoning corresponds to increasing distance in miles from the
source, and also to the presence of local production centres in Kent. The
occurrence of 2—3 types in East Sussex sites perhaps reflects the weakness
of the local competitors, particularly in the production of flagons and
mortaria (Green 1980). The distribution of late Alice Holt types presents a
more complex pattern, in which no one trend appears to predominate. The
greatest numbers of types occur in the Darent-Medway area and on Saxon Shore
forts (Richborough and Pevensey).
|
|
Sites nearer to the source (on the Surrey-Hampshire border) contain smaller
numbers of types. The pattern is complicated by the presence of local kilns
apparently producing similar wares that may not always be differentiated
from Alice Holt products, in east Kent and East Sussex (6. VIII). This
analysis bears interesting comparison with that published by Lyne and
Jefferies (1979, fig. 47) using the number of rims of Alice Holt ware as a
proportion of the total number of rims on sites throughout southern England.
The latter shows Kent sites to have a small proportion of the ware in
comparison with sites in Sussex, Surrey and the middle Thames valley, but
too few sites within the area of Kent are included to enable internal
variation to be determined with any assurance.
Two further methods of spatial analysis discussed by Hodder and
Orton (1976) may be mentioned briefly. Trend surface analysis, the
particular value of which is considered to be that ‘it allows regional
trends to be distinguished from more local ones’ (ibid., 174),
would seem to be rendered invalid by the shortage of quantified data and the
restriction of the majority of sites to a narrow strip along the North Downs
and Thames. Moreover, its usefulness as a method of predicting unknown
production centres (assuming a single centre to be involved) is in question (ibid.,
166). The second method, involving the use of gravity models to examine,
for example, marketing hypotheses, has been widely employed by Hodder
(1974a; 1974e; Hodder and Orton 1976, 189—95). The underlying concept is
one of interaction between centres (e.g. potteries, towns); hypothetical
marketing areas are constructed using Reilly’s breaking-point formula
(Hodder and Orton 1976, 188), the acreage enclosed by town walls or the
overall area of pottery kilns of a particular group commonly providing the
basic statistics. These marketing areas may then be compared with pottery
distributions in order to observe whether there is any correlation
|