Although archivists have engaged in analog information retrieval (IR) for many decades, automated archival IR is relatively unexamined in either IR or archival literature. The strength of primary source materials lies in their ability to support inference and deduction through the provision of evidence, through both content and context of documents. The hierarchical and spatial structures of archival records may be leveraged in the search for retrieving this kind of relational evidence. The bulk of work done on archival IR has used term matching as the means of retrieval. This is inadequate for research that is concerned with the relationships between documents rather than only the content of documents. This paper explores the possibilities for archival information retrieval that is based not on discrete term-searching but on retrieving evidential relationships between documents. Existing IR models are examined for their compatibility with archival IR.