Smoke alarms are one of the most promoted prevention strategies to reduce residential fire deaths, and they can reduce residential fire deaths by half. Smoke alarm function can be measured by two tests: the smoke alarm button test and the chemical smoke test. Using results from a randomized trial of smoke alarms, we compared smoke alarm response to the button test and the smoke test. The smoke alarms found in the study homes at baseline were tested, as well as study alarms placed into homes as part of the randomized trial. Study alarms were tested at 12 and 42 months postinstallation. The proportion of alarms that passed the button test but not the smoke test ranged from 0.5 to 5.8% of alarms; this result was found most frequently among ionization alarms with zinc or alkaline batteries. These alarms would indicate to the owner (through the button test) that the smoke alarm was working, but the alarm would not actually respond in the case of a fire (as demonstrated by failing the smoke test). The proportion of alarms that passed the smoke test but not the button test ranged from 1.0 to 3.0%. These alarms would appear nonfunctional to the owner (because the button test failed), even though the alarm would operate in response to a fire (as demonstrated by passing the smoke test). The general public is not aware of the potential for inaccuracy in smoke alarm tests, and burn professionals can advocate for enhanced testing methods. The optimal test to determine smoke alarm function is the chemical smoke test.