4.7 Ordinal or nominal level of information

An alternative to the reduction to ordinal or nominal data may be illustrated by an anecdote about everyday archaeological field practice. For many professional and amateur archaeologists finding, for example, a flake is not a very memorable event: these are common, after all. Finding a tool is more 'eventful' and finding an arrowhead is extensively discussed with fellow archaeologists over a beer on Monday. This example demonstrates that there are a number of find categories that are much rarer than other types and numerically attract relatively little attention. Tools and arrowheads are rarer but changes in their numbers are archaeologically very important. Site-typologically this means that on the one hand there is a group of artefact types where 2 or 3 new finds should signify an essential typological change. On the other hand there are artefact types where 20 or 30 new finds hardly affect the typological interpretation. We now propose to reduce the absolute counts to a number of ordinal classes, with class intervals increasing with the absolute number ( 'progressive' classification).

Table 10 Progressive ordinal class values

Using site 52B-168 as an example, this would result in the following ordinal classification: macrolithic artefacts = class value 1, scrapers = class value 1 and other artefacts = class value 2.

Using a progressive ordinal classification has two advantages:

1. There is sufficient stability in the data. One or two visits to a field will not greatly influence the class value of an artefact type. Of course a single find may be enough to cross a class border, but it will then take relatively many finds from the same category to increase the class value by one more once again.
2. We feel that the differences between the artefact types are expressed in a more 'archaeological way'. For instance, the absolute difference between 1 macrolithic artefact and 2 scrapers on site 52B-168 has been negated. Both artefact types occur 'accidentally'. Due to the increasing class interval the influence of some 30 new flakes on the artefact composition of the site is in a way equal to the influence of 3 new scrapers.

The interval classes can be adjusted, depending on the total size of the sites in a research area. Here we have chosen a distribution into 10 classes to accommodate slightly over 1000 finds. Arithmetically this was easy to accomplish by taking the rounded off whole value of the cube root of the number of artefacts.

Reducing the counts to these 'progressive' classes fits in well, we feel, with archaeological practice. The limitations of surface assemblages for site-typological purposes is taken account of in the analysis. No longer will the difference between 14% and 29% influence the interpretation, but only the fact that both artefact types occur accidentally.