Several research groups have addressed the problem of retrieving vector graphics. This work has, however, focused either on domain-dependent areas [1,2] or was based on very simple graphics languages. Here we take a fresh look at the issue of graphics retrieval in general and in particular at the tasks which retrieval systems must support. The paper presents a series of case studies which explored the needs of professionals in the hope that these needs can help direct future graphics IR research. Suggested modelling techniques for some of the graphic collections are also presented.