The classification of pottery as found in many archaeological sites
around the world is a time consuming and often tedious task which can
take months and even years for a single site.  Many of the seemingly
less interesting artifacts simply disappear into storage unclassified.
Automating the classification of pottery would hence be a welcome help
for many archaeologists, freeing them from routine work.  In addition
the routine classification of all finds across several sites might
also yield new and interesting results which would not have been
possible using traditional approaches.  This paper describes a first
proof of concept for the classification machinery inside such a system
and demonstrates that even with a rather haphazardly chosen set of
features and an untrained system reasonable classification results are
possible.