Recent cross-sectional studies have shown that the inability to suppress fear under safe conditions is a key problem in people with posttraumatic stress disorder (PTSD). The current longitudinal study examined whether individual differences in fear inhibition predict the persistence of PTSD symptoms. Approximately 2 months after deployment to Afghanistan, 144 trauma-exposed Dutch soldiers were administered a conditional discrimination task (AX+/BX-). In this paradigm, A, B, and X are neutral stimuli. X combined with A is paired with a shock (AX+ trials); X combined with B is not (BX- trials). Fear inhibition was measured (AB trials). Startle electromyogram responses and shock expectancy ratings were recorded. PTSD symptoms were measured at 2 months and at 9 months after deployment. Results showed that greater startle responses during AB trials in individuals who discriminated between danger (AX+) and safety (BX-) during conditioning, predicted higher PTSD symptoms at 2 months and 9 months post-deployment. The predictive effect at 9 months remained significant after controlling for critical incidents during previous deployments and PTSD symptoms at 2 months. Responses to AX+ or BX- trials, or discrimination learning (AX+ minus BX-) did not predict PTSD symptoms. It is concluded that impaired fear inhibition learning seems to be involved in the persistence of PTSD symptoms. © 2013 Elsevier Ltd.