Thirty years ago, my first foray into the theory of pottery quantification (Orton 1975) was based on two premises about the aims of quantification:
- the aim is to estimate the composition of an assemblage in such a way that it can be reliably compared to the composition of other assemblages,
- The aim is to estimate compositions of original (parent) assemblages, rather than to describe the compositions of current (excavated) assemblages.
The main theoretical problem foreseen was the testing for statistical significance of observed differences between assemblages. This was resolved by the use of the pseudo-count transformation (Orton and Tyers 1990), which is included in the software package Pie-slice (Orton and Tyers 1993). Debate has focussed on practical issues (e.g. how much time does it take?), or pragmatic outcomes (e.g. do different measures appear to ‘tell the same story’?). There has been little theory-based discussion, particularly of the premises outlined above.
Archaeologists may no longer be as interested in the quantification of ceramic assemblages as they were in the 1970s, or it may be that they do not care about the quality of data (in the sense of them being fit for purpose). It is ironic that at a time when excavation is producing many ceramic assemblages, there seems to be little attempt to exploit them by comparative studies. The way ahead may be through less abstract statistical approaches, such as the bootstrap.