Page 54 - Kennemerland VOC ship, 1664 - Published Reports

NAUTICAL ARCHAEOLOGY, 4.2
such systems have their disadvantages in that they can obstruct the process of exca- vation itself, especially where an airlift or a dredge is being used. However, since such a system removes any ambiguity in assigning individual artefacts to particular squares, and produces a high degree of relative accuracy (which is more important to us than precise locations), the advantages would appear to be overwhelming. Metre squares would be the smallest areas practical in such circumstances. The one caveat to be noted is that when a computerized recording system is being used (see below), no rigid framework is required, and an absolute accuracy of the nearest 0.25 m should be possible, thus allowing analysis of half-metre squares. It is very unlikely that any significant patterning would be obscured at this level; future experience may in fact show that analysis of metre squares is adequate.
The similarity matrices, and especially Fig. 8, contain the information on which the archaeological discussion in the next section will be founded. However, there are a few further operations which might be undertaken to assist this discussion. It would be possible to run some significance tests on the material at various stages, in order to discriminate between statistically significant and in- significant variations. However, in the present context. such a procedure seems to involve a lot of work for very little gain in information, since there is no necessary equation between statistical and archaeological significance. Consequently, the criteria of significance in the next section will be solely archaeological. Another operation to which this material seems to lend itself is clustering, which is a technique for summarizing diagrammatically the major interrelationships represented in a
matrix; in so far as this will assist in the understanding of the information contained in these matrices, it is a worthwhile exercise (Cowgill, 1968; Hodson, 1969). The method used here is a manual double-link clustering technique suitable for a small number of classes, such as we have here, whose applica- tion to archaeological data was first suggested by Renfrew & Sterud ( 1 969). The reader must refer to their paper for the details of the operation, but in essence the method involves
extracting the two highest values in each column of a complete matrix (i.e. one with the top right-hand section filled in), and then arranging the classes in a diagram so that every one of these relationships is displayed. The results of undertaking this process on Figs 5 and 8 are displayed in Fig. 9, which thus serves as a simple expression of the differing results of the two approaches. As will be emphasized in the next section, such a clustering does not represent all the signi- ficant data in a matrix (archaeologically speaking), but it does display clearly the interconnectedness of groups of variables, information which is somewhat hidden in a matrix. The fact that both diagrams in Fig. 9 indicate such well defined clusters shows that there is considerable interrelated patterning between classes on this site; if this were not the case, then the application of the Renfrew/ Sterud procedure would simply produce a confused picture.
To conclude this section, a word about the calculations undertaken above. In the appli- cation of both techniques I have used an electronic calculator, in Method 1 to convert the coefficients from vulgar fractions to decimals, and in Method 2 to transform the aggregate values to the percentages given in Fig. 7. Both of these exercises could have been undertaken manually, although to have done the percentages in this way would have been exceedingly tedious. As it was, despite the fact that the second method involved less than a quarter of the number of squares used in the first method, I estimate that its applica- tion took twice as long overall, because the calculation of the coefficients themselves is
inevitably a long-winded process. If I were undertaking this exercise without a calculator, I would limit myself to Method 1 : however, with the recent spread in the ownership of calculators, it is likely that most operators will have access to one, in which case, Method 2 is worth applying whenever the data allows it.
Archaeological interpretation
It will be recalled that our principal concern in the present study has been to try and
IS4