Acute episodes of animal mortality had severe impacts on past societies, with lasting economical, sanitary and social consequences. They remain however largely undocumented, in particular in archaeology, where the millions of animal victims described in historical texts appear to have left no material evidence. This dissertation thus strives to explore this discrepancy and to demonstrate that not only is it possible to identify archaeological deposits of mass animal mortality, but that it is also possible to diagnose the specific cause of the deaths. Theses aims are addressed by an interdisciplinary approach, drawing on methods and data from the fields of archaeology, history and veterinary science, and leading to the development of two methodological tools. The first tool allows the identification of animal mass mortality deposits through a flow chart based on objective archaeological criteria. The second presents as a protocol allowing the differential diagnosis and those of the main causes for animal mortalities in that time and place. Selected diagnostic hypotheses are then tested by targeted analyses. This allows us to identify fifty-one archaeological animal deposits probably resulting from mass mortality events in the bibliographical record, and to carry out a multi-criteria analysis of the resulting corpus in a diachronic and a species-specific perspective. Six ovine assemblages dating from the Roman period to the Modern period are then selected for an in-depth archaeological and osteological study. Likely diagnostic hypotheses are identified through our protocol in five cases out of six, one of which is confirmed by the paleoparasitological identification of the suspected pathogen.