The complainants’ memory of the rape is commonly the key and frequently the only evidence in the investigation and prosecution of rape allegations. Details, specificity and consistency in the victim’s recollection are central criteria that criminal justice agents – police, prosecutors and juries – use to assess the credibility of the victim account. However, memory research has shown these to be poor indicators of the accuracy of a memory. In this article we develop a conceptual model of the pathways through which normal features of the human memory result in complaints of rape dropping out of criminal justice process without a full investigation, prosecution or conviction, with a particular focus on the role of inconsistencies in the victim account. We provide initial, tentative evidence from a large, representative sample of rape complaints and discuss implications for criminal justice policy.