Reasoning about motives is a prominent part of the investigative process in Julian Barnes' Arthur and George and in Conan Doyle's novels themselves. In Arthur & George, it is the lack of motive that originally convinces Arthur of George's innocence, a conclusion that a subsequent evaluation of George's character supports. In many of the Sherlock Holmes novels, it is the discovery of a hidden motive that puts Holmes on the right track. On closer inspection though, the evidentiary value and logical structure of arguments from motive turn out to be problematic. In this paper, we use concepts from argumentation theory and computational models of agent behavior from artificial intelligence research to analyze the structure of motive-based reasoning in fictional and factual crime investigations and trials. This analysis is used to develop a theory that (a) accounts for the rationality of motive- based reasoning, (b) helps to distinguish plausible and implausible arguments from motive, and (c) to distinguish arguments based on motive from arguments based on intention and character, as well as from other, legally more problematic forms of evidentiary reasoning.