On September 13, 2008, Hurricane Ike was hurtling toward Galveston. Water was expected to rise 12 to 16 feet on the Texas island. The incoming surge would inundate communities. Swing sets and homes would vanish under the muck.

Despite the predictions, people did not seem to be packing up their cars and fleeing the island in droves. Then a forecaster in the local National Weather Service office issued a warning that had been used only once before -- just before Hurricane Katrina made landfall. "Persons not heeding evacuation orders," the statement said, "will face certain death."

The "certain death" message got picked up by CNN and was broadcast to homes on Galveston, all along the Gulf Coast and to the rest of America. "By getting media attention" says Gene Hafele, the meteorologist in charge of the Houston-Galveston National Weather Service office, "we got the locals' attention."

About 60 to 70 percent of the people in the mandatory evacuation zone left for safer ground -- according to Hafele's best estimate. Nobody knows the number for sure. They do know that 74 Texans died and more than 3,500 had to be rescued.

From decades of research into catastrophes -- hurricanes, fires, nuclear accidents -- disaster specialists have come to realize that people follow patterns in evacuations. Some don't evacuate because they cannot. They have nowhere to go or no way to get there. This is generally a relatively small number of people, although it can be bigger in places with high poverty rates. That was the case in New Orleans when Katrina bore down on the city.

But what about everyone else? In many disasters, age is a marker more than income. Elderly people tend to stay put. In 1979, after the accident at Three Mile Island nuclear power plant in Pennsylvania, retirees and people over the age of 70 were least likely to evacuate -- regardless of how close they were to the reactor. Many have physical or health problems, making it hard for them to move around. Also, they have survived so many other crises that they often overestimate their ability to endure the next one. Before Hurricane Ike hit, the lone holdout on Surfside Beach, Texas, was a 67-year-old man named Ray Wilkinson. "I'm tired of running from these damn things," he told a reporter from the Houston Chronicle. "If it's going to get you, it's going to get you." (He survived, but later admitted he had been "stupid" to stay.)

Two years after Katrina, 5,000 people who live in high-risk hurricane zones in eight states participated in a survey by the Harvard University School of Public Health. They were asked what they would do if government officials said they had to evacuate before a major hurricane. One third said flat-out that they would not leave. Their main rationale for staying was that they believed their homes were built well enough to survive the storm. Those who were adamant about staying were disproportionately elderly in this case as well: 40 percent of people age 65 and older said they would stay, compared with 28 percent of those age 18 to 64.

Given the built-in bias toward staying in place, local officials need to motivate residents to get up and go. But do they have to terrify people to make them listen?

Specialists who have spent their careers studying disaster warnings had mixed feelings about the "certain death" language. "The science does not say we need to scare people out of their minds to get their attention" says Dennis Mileti, a sociologist and the former head of the Natural Hazards Center at the University of Colorado, Boulder. The research does, however, show that warnings need to tell people precisely what they should do, when they should do it, who should do it (or not) and why.

The warnings must be clear, unambiguous and consistent -- even if the information on the ground is not. They need to come from many different channels (most people check with four or five sources before evacuating prior to a hurricane). And it is crucial that they be repeated -- again and again.

Most important, warnings need to be honest. The public does not tend to panic in the face of a crisis. To the contrary, most people become obedient when given clear direction -- on burning airplanes and soon-to-be-flooded coastlines.

People also prefer that officials err on the side of caution, says John Sorensen, an evacuation expert at Oak Ridge National Laboratory in Tennessee. If storm warnings turn out to be wrong, people do not refuse to evacuate the next time around -- provided officials explain why the first warning proved incorrect. "People are well aware of the fact that science is wishy-washy," says Mileti. "It's the scientists who think they need to make it perfect."

Most warnings are inadvertently written for emergency planners, not regular people -- and that doesn't help. For example, most people will never remember the difference between a tornado watch and a tornado warning. Likewise, people have become familiar with the ranking of hurricanes on a scale from 1 to 5, and since Hurricane Ike was a Category 2, many people disregarded it. But that scale measures only wind -- not storm surge, which is often far more dangerous.

The other challenge that exacerbates the human bias to linger is the ambient noise in our lives. "We have warnings from all directions -- about fire, about drugs," says Martin Lindstrom, author of "Buyology: Truth and Lies About Why We Buy." "We put up a filter and say, 'This won't happen to us.' So to communicate in an ordinary way will not work." These filters can be penetrated, but only if emergency communications are written to work the way the human brain works.

This year, people in harm's way will probably not hear the "certain death" warning, no matter how bad things get. After Hurricane Ike, the Weather Service decided to change its approach. "We had to be careful of not being overly deterministic," says Walt Zaleski, warning coordination meteorologist at the Weather Service's regional headquarters in Fort Worth. "There's always a possibility you're going to have somebody survive, and people will say, -Look! There you go! You told us we were going to die, and we didn't!' Then for the next storm, they won't react." Now, people living in high-risk areas will be warned only that they "may die" if they don't evacuate.

Over the past 50 years, evacuations have, in fact, gotten much more effective. Forecasts are more accurate and timely. Warnings are more vivid, as are storm graphics. TV and new technology such as reverse 911 calls can distribute warnings much more efficiently -- whatever the warnings may say.

Those systems will be tested again. So far, 2009 is predicted to be a busier-than-average hurricane season, with 14 named storms. Death is not yet certain.

Amanda Ripley is the author of "The Unthinkable: Who Survives When Disaster Strikes -- and Why." She can be reached at www.AmandaRipley.com.