Systemic administration of benzodiazepines before exposure to inescapable shock (IS) blocks the enhanced fear conditioning and escape learning deficits that follow exposure to IS, whereas administration before the subsequent behavioral testing eliminates the enhanced fear but not the interference with escape. The failure of benzodiazepines to reduce the IS-produced escape learning deficit when given before testing is inconsistent with a recent proposal that interference with escape is mediated by an IS-induced sensitization of dorsal raphe nucleus (DRN) activity. The present experiments demonstrate that chlordiazepoxide will block both the enhancement of fear and interference with escape responding when given before either IS or testing if microinjected in the region of the DRN. This suggests that systemic benzodiazepines fail to block escape deficits when given before testing because action at a site distant from the DRN counters the effect of the drug at the DRN.