The language used to conduct psychological tests might need be reexamined, Harvard researchers say.

The findings, published in the latest issue of the Journal of Experimental Social Psychology, revealed a significant difference in answers from the same test conducted in different languages for bilingual speakers.

Co-authored by Harvard African and African American Studies teaching fellow Oludamini D. Ogunnaike ’07, the study was based on Implicit Association Tests, which were developed as a way to determine people’s natural responses and inclination to associate an idea such as a race or religion with a positively or negatively charged adjective.

“If you ask someone ‘are you prejudiced,’ most people are going to say no. But most people do have prejudices and preferences,” Ogunnaike said.

Ogunnaike worked with Harvard pscyhology and social ethics professor Mahzarin R. Banaji to study Moroccan bilingual respondents who categorized names as either French or Arabic. They then had to repeat the experiment with the association of French as “bad” and Arabic as “good,” and vice-versa.

Ogunnaike said when the test was given in Arabic, respondents tended to more quickly associate “good” with Arabic names; in French, the results were the opposite.

He also conducted this experiment in the United States with Spanish-speaking Americans and found the same outcome: that results differed based on the language of the test.

The idea for this research stemmed from Ogunnaike’s senior thesis for a joint concentration in African Studies and Cognitive Neuroscience, for which Ogunnaike administered the IAT test in French to Senegalese respondents, but decided to change the language to the participants’ native Wolof.

“I began to wonder what would happen,” Ogunnaike said, “if the results we would get would be different if we put the test in a different language.”

Ogunnaike, who also speaks French, Yoruba, and Arabic, said he hopes his test will change the methods other researchers use to conduct surveys internationally.

“For the world of cognitive science,” he said, “[the study] means we have to be very careful about the languages that we operate in. It’s not enough just to limit tests to European languages that most researchers are familiar with. You can get completely different results.”

Ogunnaike added though the test was conducted with bilinguals, it might also apply to multilinguals, and he would like to look further at whether results would differ in languages that respondents learned after childhood.

“That would be an interesting avenue for future research,” he said.