A method for self-testing notification appliances in an alarm system by determining whether the performance of the notification appliances has degraded over time. The method may include the steps of measuring and recording initial output levels of notification features of the notification appliances at a first time, measuring test output levels of the notification features at a second time after the first time, and comparing the test output levels to respective threshold output levels that are derived from the initial output levels. The method may further include recording a pass result for notification appliances whose notification features produced test output levels that exceeded the respective threshold output levels, and recording a fail result for notification appliances whose notification features produced test output levels that did not exceed the respective threshold output levels.