When Freud first put forward his relatively late hypothesis of a death drive, he did so using a rather surprising argument from evolution. However, he came to support it with a neuroanatomical argument dating back to the earliest days of psychoanalysis. In this paper I look at these two streams of argument in some detail, and conclude that neither is persuasive. My thesis is then in two parts: firstly that the death drive, as such, probably merits no future in psychoanalytic thinking; secondly that, though mistaken, Freud's discussion of the death drive was hugely productive for the future development of psychoanalysis. Perhaps the abiding interest of the ‘death drive’ lies in this apparent paradox that mistaken theory can be truly fruitful.