Virtual stampede sees faces in crowd.

How can you stop a panicked crowd from stampeding? A team of physicists has developed an unusual computer model to make some suggestions.

While other models of crowd behavior typically treat a group as a homogeneous fluid, this new simulation views each person as an individual particle. It assumes that these people-particles vary in their willingness to follow others, and it includes the pushing and rubbing forces that can generate pressures strong enough to bend steel barriers.

It's a "dramatically different and potentially far more realistic modeling approach," says civil engineer David J. Low of Heriot-Watt University in Edinburgh in a commentary on the report, both of which appear in the Sept. 28 NATURE.

Placing narrow columns in asymmetric positions in front of exits could reduce the dangerous pressure build-up in a frightened mob and prevent it from jamming the exits, suggest authors Dirk Helbing of Dresden University of Technology and Illes Farkas and Tamas Vicsek of Eotvos University in Budapest.

Their model also predicts tragic crowd behavior known to cause avoidable deaths--groups tend to herd around one or a few exits to escape a fire while neglecting other open doors. Exits must be widened to account for this herding, Helbing suggests. Other models have assumed the flow distributes evenly over all doorways.

Helbing compares a frightened crowd to a society making its way out of an unprecedented crisis. His group's model predicts an optimal ratio of followers to independent solution-seekers.

"The methodology is promising," says physicist Bernardo A. Huberman of the Xerox Palo Alto Research Center in California. However, because there's little data on actual crowd behavior for testing the new model, he says it's premature to extrapolate the simulation results to real-world situations.

The new model also neglects important factors such as verbal communication and hand-holding, notes sociologist Dennis E. Wenger of Texas A&M University in College Station. "And it assumes people are all moving at the same velocity, which is false." Nor do the simulations account for relationships. People who know each other are more likely to cooperate in a crisis, says Wenger.

Despite its shortcomings, the particle-based approach of the new computer model represents the future for simulating everything from immune systems to ecological systems, says Kai Nagel of the Swiss Federal Institute of Technology in Zurich.