Control-based accounts of moral responsibility face a familiar problem. There are some actions which look like obvious cases of responsibility but which appear equally obviously to lack the requisite control. Drunk-driving cases are canonical instances. The familiar solution to this problem is to appeal to tracing. Though the drunk driver isn't in control at the time of the crash, this is because he previously drank to excess, an action over which he did plausibly exercise the requisite control. Tracing seeks to show that an agent's responsibility for some outcome (over which he lacked control) can be traced back to a prior exercise of control which caused (in the right way) the later lack of control. These and related cases have led many theorists to treat tracing as an indispensable component of any adequate theory of responsibility. This paper argues that tracing is in fact dispensable. I offer two strategies for explaining responsibility in drunk-driving cases (and those with a similar structure): responsibility can either be exhaustively modeled on recklessness, or exhaustively modeled on negligence. Neither explanation, however, relies on tracing. If I'm right, the case for tracing is seriously weakened.