If the future of entertainment is interactive media, some minorities are still headed back to the past.
The first comprehensive survey of video game characters, encompassing the top 150 games in a year across nine platforms and all rating levels, and weighted by each title's popularity, shows that the video game industry does no better than television in representing American society.
In some cases, video games do worse, said study leader Dmitri Williams, a social psychologist and assistant professor at the USC Annenberg School for Communication.
In his study, Williams cited research showing Latinos are making modest gains on television.
By contrast, fewer than 3 percent of video game characters were recognizably Hispanic, and all of them were non-playable, background characters.
Imagine if no Latino on television had a speaking part.
"Latino children play more video games than white children. And they're really not able to play themselves," Williams said. "For identity formation, that's a problem. And for generating interest in technology, it may place underrepresented groups behind the curve.
"Ironically, they may even be less likely to become game makers themselves, helping to perpetuate the cycle. Many have suggested that games function as crucial gatekeepers for interest in science, technology, engineering and math."
Women, Native Americans, children and the elderly also were underrepresented. For example, only 10 percent of playable characters surveyed were female, though women now make up 40 percent of video game players.
African-Americans appeared in proportion to their numbers in the real world, but mainly in sports games and in titles that reinforce stereotypes, such as 50 Cent Bulletproof.
Males, whites and adults were overrepresented.
Williams noted that some newer games give players more options for customizing their characters. Those games were included in the survey, with characters chosen randomly.
The fact that random selection did not have a major impact on the results suggests that when players have a choice, their range of options is limited.
The study itself was limited in two important ways. Many games feature non-human characters, and many are first-person games where the player never sees himself or herself. The study only included visible characters that were clearly human.
Still, the breadth of the census and the growing popularity of video games make the findings especially relevant. Total video game revenues now exceed box office and video rental receipts, Williams noted.
"In television, it was always a landmark moment when some minority or disenfranchised group appeared on the screen for the first time," Williams said.
"That kind of visibility is really the first step toward leading to public consciousness and equal treatment. These cultural markers matter."
In their study, the authors discuss possible reasons for their findings. But Williams cautioned against jumping to conclusions. "The characters the developers put in the games do not match the real world," he said. "Our thoughts about why are all informed guesses."
He did have a word of advice for game developers.
"These are highly underserved groups. It's a missed sales opportunity."
The study, titled "The virtual census: representations of gender, race and age in video games," is available online in the August issue of the journal New Media & Society, at http://nms.
For more discussion about the study, see Williams' blog post at http://terranova.
The survey covered the majority of titles sold in the 12-month period between March 2005 and February 2006 across console, handheld, PC and online platforms.
Williams' co-authors were Nicole Martins of Indiana University, Mia Consalvo of Ohio State University and James Ivory of Virginia Polytechnic Institute and State University.
Williams received funding for the survey from the University of Illinois at Urbana-Champaign, where he started the research before moving to USC.
Working at the intersection of sociology, psychology, communication and new technology, Williams studies large-scale game worlds for insights about human behavior that might be difficult to obtain in the real world.