The study used qualitative methodology in sampling and data analysis (Strauss and Corbin, 1990). After the formulation of research questions, the researchers searched for answers from a database of 120 in-depth qualitative interviews that were collected by the authors for a larger study that dealt with the experience of violence among cohabitant couples who remained together in spite of the violence. A theoretical sample of 25 interviews was selected. The findings show that two kinds of memories are stored in the archive: one raw, the other molded. Only molded memories are in use during the process of escalation. The violent man manages the archive as a "credit line" for both himself and his partner. The woman may be allowed various levels of credit, which tend to be rigid and include the extreme option of not being allowed any credit at all. In contrast, the man always allows himself a flexible credit line. Theoretical and practical implications for assessing the risk of violence and subsequent societal reaction are discussed. Possible directions for future research are suggested. 2 figures and 24 references