Within the past few decades, various proposals have been made about marking of objects in Old Japanese (e.g., Matsunaga 1983, Motohashi 1989, Yanagida 2006, Kuroda 2008, Yanagida & Whitman 2009, Wrona & Frellesvig 2010, Kinsui 2011, Miyagawa 2012), but there is still no consensus about the exact circumstances determining when direct objects are bare or accusative case marked in Old Japanese. We use the material in the Oxford Corpus of Old Japanese to examine in detail the distribution of bare and accusative case marked objects in Old Japanese texts and show that Old Japanese had ‘differential object marking (DOM)’ associated with a specific/non-specific distinction (Yanagida & Whitman 2009). Thus, in Old Japanese, accusative case marked objects are specific, but bare objects are non-specific. This paper briefly discusses cases in which accusative case is dropped from specific objects.