The traditional task in information retrieval is to find documents from a large corpus that are relevant to a query. In this project, we address a related, but different task: answering tatistical questions about a corpus. The method is designed to address queries that cannot be answered in some simple statistical fashion. For example, in earlier work we analyzed textual medical records, asking the question: "What percentage of the patients are female?"[1] This question could not be answered easily because there was no gender field in the record format.