Train station experiment reveals one way to counteract bias against Muslims


An experiment conducted in German train stations involving paper cups and escaping oranges has found that people are less likely to help a woman if she appears to be Muslim — but they’re more likely to help that same woman if she somehow proves that she shares their social values.

The findings, described in the Proceedings of the National Academy of Sciences, reveal that discrimination is a somewhat fluid phenomenon that can be mitigated — within certain limits.

Nicholas Sambanis, a political scientist at the University of Pennsylvania and one of the study authors, says he has long been interested in the discrimination faced by immigrants. In his home country of Greece, he watched as two waves of immigration in the 1980s and 1990s led to conflict in what was once a very ethnically homogeneous country.


“It’s a common argument, mainly by parties on the right, that immigrants are resistant to integrating,” Sambanis said. “They justify conflict and negative attitudes toward immigration and arguments to reduce immigration by referencing these fears that immigrants don’t want to integrate.”

But would ethnic-majority citizens feel more welcoming if they knew that immigrants were indeed adopting the cultural norms of their new countries?

To probe this question, Sambanis set up an ambitious experiment with his former colleagues Donghyun Danny Choi (now at the University of Pittsburgh) and Mathias Poertner (on his way to Texas A&M University). The work took place in 29 train stations across three German states and involved 7,142 “bystanders” who became test subjects.

The researchers chose Germany for several reasons: It has the largest immigrant population among European countries, according to a 2017 United Nations report; it’s among the most powerful countries in Europe; and it has a strong set of social norms about public behavior, which the scientists could tap into for their experiment.

German society is famous for its norm enforcement, researchers said. For example, if you leave litter lying around in Germany, there’s a good chance someone will ask you to clean it up.

With that in mind, seven teams of five people staged this scene for unsuspecting bystanders gathered at train stops:


A man at the platform would intentionally drop his used paper cup on the floor. A woman of color who appeared to be an immigrant would then ask him to pick up the cup and discard it in a nearby garbage can.

The woman’s request “signaled to bystanders that [she] shared their norms and was a civic-minded person,” the researchers explained in the study.

Moments later, her phone would ring. After she answered it, her bag would suddenly “break” and spew oranges across the platform.

At that point, the experimenters would document how many of the bystanders moved to help her gather the scattered fruit.

The scenario was repeated multiple times over several hours but varied in key details. In about half the cases, the woman would ask the litterbug to clean up; in others, that request came from another female member of the team.


The researchers also varied the orange-spiller’s appearance. The same woman of color would sometimes wear a hijab (a headscarf indicating she was Muslim), sometimes a cross (indicating she was Christian), and sometimes no religiously defined garb at all.

In some cases, the woman answered the phone in German; in other versions, she spoke in a foreign language.

Finally, in some instances, a white, German-speaking woman in secular clothing played the fruit-losing character in need of help.

The researchers performed 1,614 iterations of this two-step scene for more than 7,142 bystanders over three weeks in the summer of 2018. Then they analyzed the results.

When the orange-dropper was a white, German-speaking woman, bystanders helped her 78.3% of the time. A nonwhite “immigrant” wearing a cross or wearing only secular clothing was helped 76.4% of the time — which was not significantly different from the first scenario.

It seems that appearing to be of immigrant background does not reduce onlookers’ inclination to be helpful, at least in this particular experiment.


“It was very surprising,” Sambanis said. “It might say something about the level of multiculturalism that Germans have become accustomed to.”

But the bystanders’ helpfulness dropped if that woman appeared overtly Muslim. For instance, if the “immigrant” woman wore a headscarf, bystanders helped her only 66.3% of the time.

Acting more “German” appeared to mitigate this discrimination. The researchers found that when that Muslim woman asked a litterbug to pick up his trash, bystanders came to her aid 72.9% of the time; when she didn’t, they offered help only 60.4% of the time. That 12.5-percentage-point difference was large enough to be statistically significant, the researchers calculated.

However, a white German woman who did nothing to stop the litterbug was helped about as often (73.3%) as the Muslim woman who went out of her way to do some social good.

In other words, the Muslim woman had to work harder just to be treated the same as a white German — reminiscent of the adage that certain minority groups have to “work twice as hard to get half as far.”


To top it off, if a white German woman stepped up and told the man to clean up, the bystanders helped her the most often — a full 83.9% of the time.

The researchers also noticed big regional differences: In eastern Germany, bystanders were more likely to discriminate against the Muslim woman than were their counterparts in western Germany.

The reasons for that difference are unclear, Sambanis said. Perhaps it’s due to eastern Germany’s legacy of communism, or because economic conditions there are worse, or because residents in the east have less contact with minorities. The experiment could not discern which of these factors (if any) might have been linked to the heightened discrimination.

Donald Green, a professor of political science at Columbia University who was not involved in the study, said the experiment was “remarkable for its imaginativeness and also for the scale at which it was conducted.”

But he also pointed out a key distinction. Even though people were more likely to help a scarf-wearing Muslim woman if she engaged in a quintessentially German behavior, it didn’t necessarily affect any deeply held prejudices about Muslim women.

Those onlookers could just have been characterizing her as an exception to an underlying rule, considering her “one of the good ones” while still thinking poorly of most Muslim women who looked like her.


“At the end,” Green said, “we don’t know whether this is a prejudice-reducing intervention or whether this is simply an intervention that measures different proclivities to discriminate.”

Teasing out which of these mechanisms was motivating the bystanders’ behavior will take further study, he said.

Sambanis said he and his colleagues would continue to probe the underlying processes at work. He said he planned to do a similar experiment in Greece, where the social norms are very different from those in Germany.

“If we want to think about policy interventions to reduce these behaviors,” he said, “first we have to understand exactly what is the mechanism that causes this bias.”