Methodologies

Lost in Translation? Avoiding the Pitfalls of Multi-Lingual Research

By Craig Charney | Newsletter Clip | January 1, 2010

When it comes to multicultural or international research in languages other than English, words don’t just matter, they are everything. Even the slightest mistranslation of a single word can drastically affect the outcome of a research project.

Translation, of course, is a headache in all intercultural work, be it commercial or political.  Pepsi’s famous line, “Come alive with Pepsi,” was translated into Chinese as “Pepsi brings your ancestors back from the grave.”  Coors Brewing’s slogan “Turn it loose” translated into Spanish as “Suffer from diarrhea!”  And red faces abounded when when Jimmy Carter, arriving in Warsaw, declared, “I love coming to Poland” … and his translator used a verb implying that when the president had come, it was carnal.

But there have been equally big howlers in polling and market research.  

For instance, when one prominent market researcher did focus groups in post-communist Russia, he found a room full of very reluctant participants.  Why?  He discovered the term “group discussion” had been translated in the recruitment questionnaire as “political indoctrination session.”

Similar problems cropped up when a well-known research center planned a poll in dozens of countries.  In Ghana, the phrase “married or living with a partner” was translated into “married but has a girlfriend.”  “Separated” turned into “there’s an understanding between me and my spouse.”  In Nigeria, “American ideas and customs” became “the ideology of America and border guards”!

Here at home, a researcher looking into the digital divide was surprised that 100% of Hispanics rejected the notion that “the internet is color blind.”  Then he looked at the Spanish translation – and found it said, “The internet has red-blue-green color blindness.”  

Of course, in multi-cultural polling, numbers matter as well as words.  A political research firm was puzzled when their poll in Israel showed Arabs voting massively for conservative leader Benjamin Netanyahu – a complete reversal from prior trends.  A couple of frantic calls to Tel Aviv later, they discovered that the Arabic and Hebrew versions of the questionnaire presented Netanyahu and his Labor Party rival in reverse order – with reversed response codes too!

So how can pollsters – and their clients – avoid disasters like these?

The standard answer is “back translation”: having the questionnaire translated back into English by a second translator and comparing the result with the original.  This approach is fine for catching gross errors – like turning “customs” into “border guards” – but may not avoid subtler errors of usage.

At Charney Research, we go beyond the norm and use two additional methods on multi-lingual projects.

• Careful review of the translated questionnaire.  Often we will go over it line-by-line with supervisors and interviewers during briefing, to make sure not only that they understand it but that it makes sense and says what it is supposed to.   When this is impossible we have a bilingual individual compare it closely with the original.
• Creating bi- or tri-lingual questionnaires.  We will put together a draft with both English and foreign language versions of every question side-by-side to facilitate comparison and ensure that questions and answer codes are the same in each language.

These extra steps have kept us out of trouble – and will keep you out of it too.