Page 57 - Kennemerland VOC ship, 1664 - Published Reports

K. MUCKELROY: WRECK SITES
these classes and the clay pipes is slightly lower (153 and 146), but a difference of this magnitude cannot be significant; it is likely that all these items were stowed in the same area of the ship.
The most outstanding difference between Fig. 9A and 9B is the cluster of classes VI and XI in the latter. Reference to Fig. 7 shows that the high correlation produced by Method 2 is due to the fact that both personal possessions and bones are highly concentrated in square F3 (70.43% and 64.29% respec- tively). In the case of bones, whose presence is dependent on favourable conditions for preservation, this concentration is principally due to the pockets of fine mud found in and around this area, and nowhere else on site. To the extent that the category of personal possessions includes equally perishable items (especially leatherwork), the correlation is therefore determined by factors of preserva- tion. However, the bulk of class VI is made up of very hardy material (notably fine jewellery and pewter utensils), so that other unidentified factors must be relevant as well. On the other side, class VI is linked to the clay pipe cluster; this is an interesting relationship, since it is likely that both classes consist of a mixture of items that were in use by those on board the ship, and items of cargo. The evidence regarding clay pipes has already been
together has meant that no study has been made of the way in which the ship broke up. These, and other, points will be investigated once the whole of this site has been excavated. For the moment, the important conclusion has been reached, by methods which are both explicit and objective, that the sea-bed distributions on this site contain relevant information, evidence which it is our duty to study in more detail. One way in which this might be done is outlined in the next section.
A wider conspectus
Looking forward to the time when the excava- tion of the Kennemerland site is completed- to the extent that any excavation is ever completed-it is obvious that several types of analysis will be required. For example, it will be possible to investigate in detail the way in which the ship broke up, bearing in mind the traditional account (Bruce, 1907) and the total spread of the finds across the South Mouth. However, for the present I will mention only the possibility of under- taking a more sophisticated analysis of the relative distributions of all the finds. So far, it has been established that there are several factors, some of them of archaeological significance, lying behind the patterns re- corded on the sea-bed. The requirement now is for a technique for quantifying the con-
mentioned, while the composition of the
jewellery assemblage has already lead to
suggestions of a collection intended for tribution of these various factors; among
private trading (Price & Muckelroy, 1974: 262). It is also worth noting the relationship which achieves the lowest score of all, that between classes V and VI (8). The dissimilarity here is probably compounded of numerical inequality, differing value, and differing ship- board status; the one factor that cannot be operative is differing fabric, since both classes are predominantly metallic.
This type of analysis can only shed light on those factors which can cause patterning on the sea-bed; it can say little or nothing about factors which have tended rather to eliminate certain items from the sea-bed, factors such as differential preservation of materials, or the depredations of salvors. The study of these must proceed by other methods. The fact that the areas excavated were relatively close
those available in the statistical armoury, the most promising would appear to be some kind of factor analysis. In these techniques, the factors determining the variance in a matrix are isolated, and the ‘factor weighting’ is calculated, which expresses the proportion of the variance of any particular variable ‘explained’ by that factor; by consideration of these weightings for each factor, it should be possible to identify the nature of that factor archaeologically (Clarke, 1968: 562-4; Cowgill, 1968). It should be noted, however, that similar exercises in other archaeological contexts have experienced problems in equat- ing statistical with archaeological factors
(Doran, 1974). Nevertheless, the attempt is certainly worth making.
In view of the fact that the manipulations 187