Automotive collision warning systems (CWS) can enhance hazard identification and management. However, false alarms (FAs), which occur as a random activation of the system not corresponding to a threat and not interpretable by the driver, and unnecessary alarms (UAs), which occur in situations judged hazardous by the algorithm but not by the driver, may limit CWS effectiveness. A driving simulator was used to investigate the influence of CWS (accurate, UA, FA, none) and distraction on driver performance during non-critical and critical events. FAs and UAs differentially influenced trust and compliance. FAs diminished trust and compliance, whereas the context associated with UAs fostered trust and compliance during subsequent events. This study suggests current warning descriptions based on signal detection theory need to be expanded to represent how different types of alarms affect drivers.