When a fire alarm rings, many people think, "It's probably nothing." This is called normalcy bias. In a 1968 study, smoke came into a room. When students were alone, 75% went out to report it.

But when they were with two calm people, only one in ten reported it. In real disasters, panic is rare. The bigger problem is that people start to move too late.

In disasters, many people think "It's nothing serious" and delay escaping. This tendency is often called normalcy bias. A famous 1968 experiment showed how other people make it worse. While students filled out a questionnaire, smoke was sent into the room. Of those who were alone, 75% left to report it.

But when they sat with two people who calmly kept working, only one in ten did. Some even kept writing while rubbing their eyes. They later said they thought it was steam.

In addition, many people believe bad events are less likely to happen to them than to others. Real panic is actually rare in emergencies. The greater danger is moving too late.

Disaster specialists use the term "normalcy bias" for the tendency to downplay warning signs and assume that everything will remain as usual. A classic 1968 experiment shows how social cues can intensify this tendency. As students completed a questionnaire, smoke was piped into the room through a vent.

When they were alone, 75% left to report it, but when seated with two confederates who calmly continued working, only one in ten did so; some kept writing while rubbing their eyes and later explained that they had assumed it was steam. A related phenomenon, optimism bias, was documented in 1980, when students overall judged negative events to be less likely to happen to them than to their peers, which is statistically impossible for everyone.

Studies of survivors suggest that panic is rare, and that people frequently help one another and behave calmly. The real hazard is delay. Two caveats apply, however. The smoke study was small, with only ten to twenty-odd participants per condition, and normalcy bias is better seen as a label for several processes than as a precisely defined construct. Research on evacuation shows that people often spend time confirming warnings and checking on family before acting.