FORT WORTH, Texas - Like a Hollywood actor, the newest member of the National Weather Service can turn his emotions on and off at a moment's notice.
That makes him the ideal choice for motivating people to take cover when seconds count, meteorologists say.
"Tom," who began work last month, is the newest generation of automated voices broadcasting the National Weather Service's daily forecasts and advisories over NOAA Weather Radio. Because the voice is computer-driven, its speed and intensity can be adjusted to make severe weather warnings sound more urgent than daily forecasts.
Studies have shown that such a change increases the likelihood that people will pay attention, said Bill Bunting, chief meteorologist for the service's forecast office in Fort Worth.
"There has been a lot of research sociologically and psychologically in the area of how people perceive warning information and how they respond to it," he said.
"Tom" replaces "Donna" and "Craig," the automated voices used since last year. All three are produced by software that breaks down recordings of a real human voice into their smallest parts and reassembles them to form new words. Drawing upon 30 to 40 hours of recorded speech, the software can read aloud any text, said Robert Rieger, text-to-speech product manager for SpeechWorks International, the Boston firm that developed them.
The new voice - named for the Boston-area singer who recorded it - is a little easier to understand, Rieger said.
"It's much richer, more animated, and closer to the Holy Grail of a highly intelligible, natural-sounding voice," he said.
David Sattler, an associate professor of psychology at Western Washington University, has studied the public's response to severe weather events. He said both the delivery and wording of warnings can work together to persuade people to seek shelter.
"The intonation can convey the seriousness of the situation," he said. "But my hunch is that it's really what that automated voice says that's going to be most important, as opposed to the way it says it."
The automated voices also allow meteorologists to focus on the content of the messages being broadcast, said Walt Zelesky, warning coordination meteorologist for the weather service's southern region.
"Let's say you can save the two or three minutes that used to go into preparing for live broadcasts," he said. "If you've got a tornado moving at 60 mph, that's two or three miles of [a town] that you have saved."
The weather service's first automated voice, "Paul," began broadcasting in 1997. Its words were entirely computer-generated. At times, Zelesky said, it could be hard to understand.
"The newer voices may not always sound perfectly human, but they're about as close as you can get," he said.
They might also sound familiar, Rieger said. "Craig," "Donna" and "Tom" read e-mail aloud for America Online and Yahoo! and give driving directions for OnStar.