An outbreak of a deadly disease in a virtual world can offer insights into real life epidemics, scientists suggest.
The "corrupted blood" disease spread rapidly within the popular online World of Warcraft game, killing off thousands of players in an uncontrolled plague.
The infection raged, wreaking social chaos, despite quarantine measures.
The experience provides essential clues to how people behave in such crises, Lancet Infectious Diseases reports.
In the game, there was a real diversity of response from the players to the threat of infection, similar to those seen in real life.
Some acted selflessly, rushing to the aid of other characters even though that meant they risked infection themselves.
Others fled infected cities in an attempt to save themselves.
And some who were sick made it their mission to deliberately infect others.
Researcher Professor Nina Fefferman, from Tufts University School of Medicine, said: "Human behaviour has a big impact on disease spread. And virtual worlds offer an excellent platform for studying human behaviour.
"The players seemed to really feel they were at risk and took the threat of infection seriously, even though it was only a game."
She acknowledged that a virtual setting might encourage riskier behaviour, but said this could be estimated and allowed for when drawing conclusions.
She said a major constraint for epidemiologists studying disease dynamics at the moment was that they were limited to observational and retrospective studies.
For example, it would be unethical to release an infectious disease in real life in order to study what the consequences might be.
Computer models allow for experimentation on virtual populations without such limitations, but still rely on mathematical rules to approximate human behaviour.
A virtual world may be a way to bridge this gap, said Professor Fefferman.
Her team at Tufts are looking to use models such as the World of Warcraft to further study human behaviour, particularly in relation to disease outbreaks.
Dr Gary Smith, professor of Population Biology and Epidemiology at the University of Pennsylvania, has been working on modelling infectious diseases.
He said: "Very few mathematical models of disease transmission take host behaviour into account."
But he questioned how representative of real life a virtual model could be.
"Although the characteristics of the disease could be defined before hand, once released into the virtual world, the study is just as 'observational' as disease outbreak studies in the real world.
"Nevertheless, I suppose one could argue that the proposal describes an opportunity for study that we might not otherwise have."