You can open a free account and pose your questions and request at Proz.com (freelance professional translator forum). 

http://www.proz.com/

Some volunteers may actually offer their help to you without costs.

Of course, the better the professional credentials the authors of the questionnaires are, the better the quality the study will likely to be. it adds to confidence about the accuracy and validity of results. However, translation is an art, not really a "learned" set of skills. Therefore, professional experience, rather than formal certs, is important. I am 100% sure, the proz.com people will agree on that. 

Good luck. 

Thank you very much Martina Gaisch and Nai-Kei Wong for your valuable answers.

Second part of my question is still unanswered.

How is the sample size selected for translation-based studies like this? Inclusion and exclusion criteria and number of participants etc...

I fully support  Gaisch's  and Wong's views. Remember that translators do not necessarily have the domain specific language knowledge needed, so the quality assurance should involve having experts looking at the questions. It is just as important, however, to pilot the questionnaire, if necessary repeatedly. Linguistic accuracy does not always equal clarity in the minds of the respondents. 

Normally, to translate a questionnaire you go through the 2-step or 3-step forward-back translation method:

• At least 2 bilingual translators with original language being the one you are translating into do independent translations, and although they must be as literal as possible they should also make suggestions to make the question understandable for the culture they are going to be used in
• Translators not only translate the whole questionnaire, but add a rating of difficulty to each idem and response, and/or mention which problems they found: the correct way to say a certain thing in the new language might be much longer and not suitable to test format, etc. A typical problem is often with item answers, when gradations are not understood exactly the same, etc.
• An expert pannel (experts in questionaires, in the topic the q. is about, and if possible a person from the target popultion) goes through the translation, with a table that shows the discrepancies among translations. A final translation is achieved by choosing the best option in each case, or mixing both / the 3 translations to get a version satisfactory for the expert pannel.
• Often, you do a new independent back-translation, which is now totally literal so as to check nothing is lost in the final version.
• Sometimes the consensual translation and the back translation are sent to the original authors as a proof of good practice.

After this process, and before actually using the new version fo the questionnaire, you go through a cognitive debriefing process. A number of patients or target population answer the questionnaire alone, by proxy, or however the instrument is ment to be answered. Then, a trained interviewer goes through the whole questionnaire with this person, making it relaxed conversation in which you re-ask the items in a way to see if the answer was correct, and in discrepancies you ask if they understood this or not, and how the target would have worded it.

I'm sorry I'm rushing this answer, you can find more information in these links:

• WHO - Process of translation and adaptation of instruments: http://www.who.int/substance_abuse/research_tools/translation/en/
• A paper on the cultural adaptation into Spanish of a questionnaire:

http://www.archbronconeumol.org/en/cultural-adaptation-into-spanish-of/articulo/13133592/

http://www.who.int/substance_abuse/research_tools/translation/en/

http://www.archbronconeumol.org/en/cultural-adaptation-into-spanish-of/articulo/13133592/

Thank you for your responses. What about sample selection in these studies? Sample size, inclusion and exclusion criteria etc? 

Let's consider a scale with eight (8) items, how much would the sample size be for it? 5 participants per item or more?

Additionally, is it necessary to have target population with disease/disorder if we are testing a relevant tool or it can be tested on healthy individuals?

The target population should be the real target population.

I'm sorry, I can't really answer that but there are ways of estimating the sample size. I am not the statistician at our team, I'm one of the translators among other things.

There are many papers on the subject, I just can't search for them now, sorry.

Maybe a statistician turns up!

This depends on the type of your questionnaire, i.e. if the questions are easy and straight forward and free of linguistic and cultural mismatch, then you do not need experts.. But if  the translation is between two  incongruent languages and the questions are loaded with cultural specific items then you are obliged to seek help form an expert in the field of translation. 

As to your second question, The questionnaire is based on the size of population you are dealing with. However, this is relative and  depends on your data.

@Ramadan Elmgrab could you be more explicit in answering my second question?

It is highly recommended that the questionnaire items be clear and unambiguous to achieve and ensure content validity. But for that, you would be better off to have an expert in the field who has a full command of the target language. 

For this type of translation, one approach could be to have two individuals who have a good grasp of the subject matter and know the other person's language. This way the cross-cultural subtleties of the two languages is not lost. Translation is more about translating the meaning than the text. I do not believe that it is necessary either to have a formal education as long as they are very well grounded in their first language. They should also have a sufficient knowledge of the other language and culture as it affects the sublte differences of either language with respect to specific terminology or concepts being addressed.

I second Dr. Yaghoubi-Notash.

In the commercial questionnaire validation practice, the process is actually very costly and elaborate. It may proceed by:

(1) first defining the questionnaire items (questions, scales) in a language that is most clear in articulating the concepts (e.g. English or French);

(2) cross-validating this source text by editors and proofreaders, and pre-testing it by subjects (i.e. English or French speakers), followed by more editing and revision so that accuracy and precision can be guaranteed.

(3) after that, commissioning a professional translator (usually freelancers such as those at Proz.com in an ad with pay, or by recruiting reliable volunteers) to translate the source text into a target language. This translator is usually a native speaker of the target language;

(4) after that, commissioning another pro translator and native speaker to translate the same text.

(5) a third native speaker, usually a very experienced language expert, is tasked to "pool" (integrate) the two translated texts into a unified version.

(6) this unified translation is sent to an outsourced editor (anther native speaker, bilingual) to check (edit and proofread, etc.) the accuracy and precision of the translation, comparing items on the source text

(7) after that, the translation is to be  validated and pre-tested in local subjects who speak the target language. Local interviewers (usually one, on pay) will be appointed to recruit such local subjects. The questionnaires will be tried out by the subjects. The subjects will be asked if there are any items that are not clear to them (these constitute the basis for improving the clarity and accuracy of the translated questionnaire). The results of subject feedback will be summarized by the appointed interviewer, and sent back to the commissioning agent. 

(8) an editor (bilingual expert) recommend changes based on local subjects feedback to the translated text. 

(9) after final proofreading, it's finally done. 

***

Such expensive and labor- and capital-intensive questionnaire validation work is practiced by commercial medical survey companies in the UK. They recruit trusted freelancers at Proz.com to provide the service. 

Occasionally, volunteer's help is available, and the people at Proz.com are friendly. 

***

In commercial survey validation, the local subject sample size is typically 5, with different age and sexes. 

For academic research purpose, the sample size would be determined by normal statistics considerations. 

You may need to control for sex and age cohorts. 

Minimal sample size is 30 for normal distribution (have 40-50 to be safe). And of course, the larger the sample size the better. (However, if your questionnaire is large (over 20 items), practically, it would not be possible to have many subjects happy to give you the answers. Processing and analysis costs will also be blown up. 

I second Nakei Wong, though I think some of these steps can be compressed, otherwise ti would take forever to get the final product before it is even ready for standardization. 

My concern would be with how "brilliant" is measured (e.g., level of expertise in both L1 and L2, knowledge of the subject matter and the scope of the treatment, etc) and whether or not such bilingual translator is capable of ensuring the validity of the questionnaire, as per Dr. Yaghoubi-Notash.

My opinion is that it is better to do so. Professionals do professional things. My personal experience in translation, e.g., when translating materials that are in the field of car mechanics into another language, is that the lack of professional knowlege may render the transated text unnatural, even to the degree of confusing or misleading. 

Yes, it is important to have professionals because translation-related decisions have a direct impact on the validity of the research and its report. In general, it is important to keep in mind the following factors which affect the quality of translation: the linguistic competence of the translator/s; the translator’s knowledge of the culture of the people under study; the translators’ knowledge (even basic) of the subject matter. When I worked on a translation project many years ago, I also remember we talked a lot about a need for translators to be explicit (when translate data from one language to another) in describing their choices and decisions, translation procedures and the resources used.

Definitely! If you want to do the right thing.

As some of you point out, I would summarize that translating a questionnaire is a very delicate task. It is important to have not only a bilingual expert, but a native speaker who can differentiate dialects (within or between countries speaking the same language), from errors coming out from a bad translation. Validity can definitely be affected.

Definitely! Otherwise, the real intent of your questions may be lost in translation.

Well, well... a questionnaire is usually called to get feedback on certain issues, disregarding whether in relation with a certain product or social context or anything you imagine. I doubt that a mere translation for such a questionnaire from, say, Swedish to Urdu leads to a reasonable solution, because the social and psychological context is very different. Remember, always remember, that a language is NOT a list of words and expressions, but something a lot more complex: habits, gestures, connotations, social norms etc. 

If we limit to translating from a language A to a language B, both belonging to the same cultural context, e.g. from Swedish to Danish, things are considerably simpler, but—keeping the European frame—things become a lot more complicated if we try to translate a questionnaire from, say, Finnish to Greek. 

Therefore, a questionnaire, more than other texts, require an adaptation of terms, implications and connotations. And, yes, the translator(s) must handle a complex set of social and linguistic norms, not only to memorize a list of words.

Absolutely, Sorin. The social, political and cultural environments can be completely different. Questions asked within one culture may be irrelevant or even unacceptable in another culture. Translation is far more complex than just changing the language used and mapping words. Just think of the difference in attitudes to women between cultures, for example. Think of the difference in religions and religious attitudes. The author(s) of the questionnaire is/are quite possibly not aware of how their questions are rooted in cultural suppositions. You need someone who has a deep understanding of both the source and target cultures and the purpose and aim of the questionnaire itself - ideally the original questionnaire author(s) need to work with the translator, and it can be a very subtle process of re-creation.

Sian and Sorin are both correct.  But the political aspect also needs to be considered for some questionnaires.  For example, a questionnaire on schooling in German would need to be adjusted for Belgian, Austrian, Luxembourgian Swiss and German people because of the different school systems (and even with-in Germany for some issues).  "Yuh gotta know the territory."

Thank you very worthy colleagues for providing explicit answers to my question. 

I have understood that if you want an Questionnaire to have a valid translation, translate it from one language to the other, and then have a different person translate it back to the first. E.g.  An Urdu questionnaire translated into English needs to be checked over for Clarity in English, and then have another person translate it from English to Urdu.  You may want to do this with more than one set of translators.  

If the questionnaire comes back into Urdu from the English with inaccuracies, then you need to look at the English version and tweak it for accuracy and sociolinguistic meaning, and then try again with yet another translator.  None have to be professional but all should be familiar with the topic.   

In Grad school to took a course for translating French into English.  I wish I was good at going the other way!

@Joan Plonski. The essence of the whole discussion is that a questionnaire is NOT a typical translation issue, as it refers to a lot of extralinguistic, social and psychological issues, not mainly to language.

It depends. Some years ago, when I was Director of the Center for Applied Linguistics in Washington, D.C., the U.S. Census Bureau decided to prepare a Spanish version of the census questionnaire, and piloted a preliminary version. I don't know who the original translator was, but many agencies just assume, first,  that because some staff member speaks a language natively, they are qualified to translate a document, and secondly, that all languages are monolithic. A friend from Texas who had received the Spanish questionnaire contacted me to say that she did not know some of the terms used. I got hold of the questionnaire, and got together a committee of CAL staff from Texas, California, Puerto Rico, and Cuba, who went over the QR and produced a "consensus translation", which included alternative terms used in these different areas. We then piloted our version in various communities across the U.S. and made some additions based on our fieldwork. The terms which showed the most variation were not technical terms known only to specialists, but ordinary everyday terms. One that I recall was the translation of "factory". A common term in Mexico, and hence among Mexican Americans, is "fabrica", while Puerto Ricans used "planta". At least two or three other regional terms were used, so all of the alternatives were incorporated in the final version.

Just because someone is educated in a language does not mean that they are sensitive to all of the different terms used by ordinary people, since in many countries, "education" is equivalent to alienation from traditional culture. Some years ago, academics in an African country compiled a dictionary of automotive terms for a major national language, but some linguists found later that  the compilers had in many cases made up words based on logical derivation, but no one had bothered to check with local automobile repair men to find out what terms they were actually using. To be valid, a questionnaire should, as Alejandra said, be grounded in actual usage, and should never be left up to one person, no matter how well informed or brilliant, to translate.

Further, it should be at least experimentally piloted among typical target respondents, to get live feedback on intelligibility. Finally, the circumstances for the use of the questionnaire should be carefully examined: the characteristics of the interviewer and the context of the interview can sometimes seriously distort responses. The administration of a questionnaire is a real-time social event, and can be significantly affected by social conditions.

     --Rudy Troike

One method used for this type of work with language oriented questionaiires and tests (for language evaluation, for ex.) is to back-translate. That is, once you have your forward translation (Language A to Language B), you ask a second person/professional to translate B to A. You can check for coherence between the original document (A) and the new backtranslated one (A-prime). If is is not coherent, you can go back to the B version and modify it.

Thanks everybody for such informative and important feedback.

Let's come on its testing and analysis part now.

Pilot testing steps include recruiting the target group and administering finalized version of translated (B) questionnaire on them. Subjects are then asked whether they had any difficulty answering the questions, whether they found any of the questions confusing, upsetting or containing difficult vocabulary. If a subject reports having trouble answering any of the questions, he\she is asked to suggest an alternative way to formulate it (provided the subjects do not discuss items with each other while they  complete the questionnaire). After analyzing the patient comments and making minor changes in the language-if necessary, the final translation is obtained and approved.

Correct me if I'm going wrong in any of the step.

For analysis part, I would need guidance from the respected researchers around the globe.

Yes @Arif Jawaid the questionnaires will be administered face to face.

For the analysis part, I need suggestions from the researchers too.

I would like to emphasize that the nature of face-to-face interviews implies a heavily fraught social and cultural relationship. The sex of the interviewer vis-a-vis the interviewee, the social status, ethnicity, or age of the interviewer versus the interviewee and even the history of the relationship between ethnic groups can strongly affect the outcomes of an interview. You have not provided much information on the context or content of the questionnaire, nor of the identity of the interviewers or interviewees, so it is difficult to comment specifically on these issues. Members of subordinate social and ethnic groups are likely to give answers that conform to expectations, or to be reticent to respond, to questionnaires administered by more educated or ethnically different interviewers. The issue of the language and even the regional or social dialect(s) of the interlocutors can be critical, as I discussed before. For our responses to be helpful, the whole purpose, context, content, and plan for administration of the questionnaire need to be made clear.

Thanks everybody for such informative and important feedback.

Let's come on its testing and analysis part now.

Pilot testing steps include recruiting the target group and administering finalized version of translated (B) questionnaire on them. Subjects are then asked whether they had any difficulty answering the questions, whether they found any of the questions confusing, upsetting or containing difficult vocabulary. If a subject reports having trouble answering any of the questions, he\she is asked to suggest an alternative way to formulate it (provided the subjects do not discuss items with each other while they  complete the questionnaire). After analyzing the patient comments and making minor changes in the language-if necessary, the final translation is obtained and approved.

Correct me if I'm going wrong in any of the step.

For analysis part, I would need guidance from the respected researchers around the globe.

I'm unsure why you have chosen to rely heavily on a qualitative analysis of the survey instrument. Since this is a quantitative survey, I'm assuming that you have constructed multi-item scales, and that these scales have been analyzed to ensure that they offer a robust measurement of the scale's construct. If evidence (i.e., alpha levels) suggests that certain items do not contribute to the overall scale reliability, then items should be excluded. This is the primary aim of piloting a questionnaire.  

I have seen too many instances where highly sophisticated statistical analyses have produced self-verifying false or vacuous findings. If you've watched the British TV program "Yes, Minister", you may recall a famous episode where Sir Humphrey tells the minister that he can get whatever results he wants to a poll on gun control by the questions he poses in the questionnaire. If the construct is wrong, the most elaborate statistics in the world will not make it right, but people who believe in the power of statistics will believe the results.

Yes! I also remember the "under consideration" vs. "under active consideration" joke. So, while I concede Rudy's point, my comment was in reference to a scholar using an existing scale in a new setting or with modifications. We would hope that the scale was developed meaningfully and that item wording reflects the appropriate construct. However, following from that, simply taking it for granted that the modified scale still has good measurement properties without testing the reliability coefficients may lead to unnecessary grief in later data analyses simply because the scale is not a robust measurement of what you are trying to measure.  

I have not watched the show but I do agree with @Rudy Troike and  @Philip Hiver . 

Please guide me on role of Cronbach's alpha in these studies?

 In order to have a good and correct translation, I think that the person translating the questionnaire should have a formal education, be fluent in both languages and also be familiar with both cultures.  

This could be considered as "necessary but not sufficient" some of my PhD students have had work translated by professional translation bureau, who have provided excellent samples of prior work but done a pretty shoddy job on the work (and charged a lot for it). In my experience they key to a good questionnaire is a good pilot test of the experimental procedure and materials. Get native speakers of the target language, as close to your demographic groups as possible,  to comment on whether the questionnaire makes sense.If you want an example of a study (in English - semantic similarity) that pays very careful attention to materials selection, population sampling, questionnaire measurement scales and permissible statistical techniques, look at: https://www.researchgate.net/publication/262411640_A_new_benchmark_dataset_with_production_methodology_for_short_text_semantic_similarity_algorithms

Article A new benchmark dataset with production methodology for shor...

Yes it is necessary as you want to make sure that the bilingual experts who function as translators can translate and adapt items of the questionnaire accurately to reflect their original meaning and ensure the concept is relevant and appropriate for the target population. The translators should be from different backgrounds.

I think you were referring to sample size of psychometric testing of the questionnaire (reliability and validity testing). It will depend on types of test you use. For example for exploratory factor analysis, you need minimum of five per item. If your questionnaire has 20 items, it means you need at least 100 participants or subjects.

Nice Dear James Dominic O'Shea