In the SPICE project we are building a distributed catalogue of life, which will eventually be formed from up to 200 autonomous taxonomic databases. We are faced with a number of challenges, which include the scalability of the system; the accommodation of partial or missing data; queries which are potentially very expensive computationally, where it is difficult to determine which databases will contain data matching the queries, and the effective integration of heterogeneous databases at the knowledge level. In this paper we present the architecture on which SPICE is being built, and we explain how, within our SPICE architecture, we will be able to explore and develop new techniques to enhance access to the SPICE distributed database.