In a new study, 221 college students participated in an online chat room in which they watched a fellow student get "bullied" right before their eyes.

Only 10 percent of the students who noticed the abuse directly intervened, either by confronting the bully online or helping the victim.

The abuse wasn't real - the bully and the victim were part of the experiment - but the participants didn't know that.

"The results didn't surprise me," said Kelly Dillon, lead author of the study and a doctoral student in communication at The Ohio State University.

"Many other studies have shown bystanders are reluctant to get involved when they see bullying. The results disappointed me, as a human, but they didn't surprise me as a scientist."

The bright spot in the results was that a much greater percentage of participants who noticed the bullying (nearly 70 percent) indirectly intervened by giving the bully or the chat room a bad review when given the opportunity later.

"Most of the people didn't stand up to the bully, but behind the scenes they did judge the bully harshly and try to pass that information on later when the incident was over," Dillon said.

Dillon conducted the study with Brad Bushman, professor of communication and psychology at Ohio State. Their results appear in the April 2015 issue of the journal Computers in Human Behavior.

For the study, the undergraduate students were led to believe they would be testing an online chat support feature that was part of a server used for online research surveys and studies.

The chat monitor was available to provide assistance while participants completed a series of personality surveys. The chat room window was visible on the computer screen while the participants completed their surveys. Participants were told they would be asked to evaluate the chat monitor at the end of the study.

Three minutes after the participants began taking their surveys, the cyberbullying of the victim began, following a script written by the researchers. Participants could see in the chat window that the victim was having trouble saving a response in the survey.

A conversation ensues, with the chat monitor becoming increasingly hostile toward the victim.

"We had the chat monitor say things like 'How did you get into college if you can't even take a survey?'" Dillon said. "Finally, after getting increasingly aggressive, the chat monitor tells the victim, 'Figure it out yourself.'"

After three minutes had passed, the victim asked another question and the scripted abuse began again. In the script, the victim did not respond to the rudeness at all.

About 68 percent of participants said later that they noticed the cyberbullying in the chat window. Of the one in 10 who noticed the abuse and responded directly, more than half (58 percent) reprimanded the bully. One response, for example, was "How are you being helpful at all right now?" A quarter of those who responded insulted the bully, saying things like "I can smell the odor of loser from you."

Less commonly, some participants offered technical support and social support to the cybervictim. One person complimented the victim, saying, "I'm sure you're smart!! You'll get it."

After filling out their surveys and testing the chat room, all participants were asked to grade the chat monitor and indicated whether they would recommend the chat room function to future participants. Nearly 70 percent of the people who noticed the cyberbullying and who didn't respond directly to the abuser gave bad marks to the chat monitor and/or didn't recommend use of the chat room - both of which were classified as indirect intervention.

That left about 15 percent of the participants who noticed the cyberbullying and who didn't directly or indirectly intervene.

Dillon said we shouldn't judge the people who didn't intervene too harshly, because we don't know why they didn't respond.

"At the end of the study, when we told participants about the true purpose of the study, many who didn't respond or who responded indirectly said that they wished they had directly intervened. Many said they wanted to respond to the bullying, but weren't sure what they should do," Dillon said.

"We all do that occasionally. We're all bystanders at some point."

Dillon said this research may aid in designing interventions that can help bystanders find ways to stop cyberbullying. For example, this study showed that relatively few participants responded directly to the victim, which may be most helpful in some cases.

"If witnesses think that they have to confront the bully, that may be tough for many people to do. But this study shows how they can help the victim, or remove the victim from the situation. That may be the best strategy in some cases," she said.