Translation of survey instruments Alisú Schoua-Glusberg, Ph.D. Research Support Services DC/AAPOR Presentation – 7/7/04

Research Support Services Translation methods for survey instruments This presentation will focus on: problems involved in questionnaire translation methods and approaches for the translation of survey instruments, use of qualitative methods to assess translation quality and the performance of the translated instrument, providing survey translators the information they need.

Research Support Services When is Questionnaire Translation Necessary? Cross-national studies Cross-cultural studies National studies by Federal Statistical Agencies Studies of special populations Studies where sample falls in areas with significant concentrations of speakers of other languages

Research Support Services Factors to Consider to Define the Target Audience of Translated Instruments Language minorities Immigrant populations Monolingual (or at least not bilingual) Language different from designers’ Lower level of education Age at immigration Hybrid culture Different degrees of acculturation

Research Support Services Source & Target Question Characteristics to Consider Meaning Style Complexity Source flexibility Cultural aspects Existence of equivalent realities

Research Support Services Using Previous Translations of Questions vs. Translating Anew Issues: Maintaining longitudinal comparability vs. improving questions Is priority to compare with past translations or with English questionnaire?

Research Support Services Questionnaires are a Complex Text Type Little context to clarify ‘intended meaning’ Ambiguous Measurement properties lead to ‘surveyspeak’ and ‘scalespeak’ ( Harkness 1996 ) Translators thus often not versed in questionnaire discourse ‘rules’ Special kind of conversation

Research Support Services Question ‘Equivalence’ is Expected in Translation Expectation that the translated question… Says the same thing Means the same thing Measures the same thing Measures equivalently (e.g. scales) Imposes same burden on respondents Meets reliability and validity requirements

Research Support Services Translators’ Job Translators make decisions about: Semantics Syntax (structure) Pragmatics They need to understand: Intended “meaning” in order to translate Covert (measurement) intention and requirements

Research Support Services Translators therefore… Have little context and co-text Have little supporting documentation Often have no questionnaire ‘author’ to consult May experience uncertainly about what to “match” May make decisions based on their experience with other types of text

Research Support Services Risks of Overly Close Translations Focus on meaning of words rather than meaning of questions Rs are inadvertently asked a different question Processing is more complex Translated questionnaire sounds unidiomatic

Research Support Services Steps in producing and testing quex translations Translation Translation review Decisions/Adjudication Quality control Qualitative research Pretesting Documentation

Research Support Services Survey Translation Approaches One translator- one translation (direct translation) Multiple translators – one translation (split committee) Multiple translators – multiple parallel translations (parallel committee)

Research Support Services Committee Approach Three translators prepare translations independently (split or parallel) Reconciliation meeting with referee Qualitative research with monolinguals (focus groups and/or cognitive interviews)

Research Support Services Committee Approach: Reconciliation Meeting Question-by-question review Reaching consensus when possible Providing alternatives if no consensus possible Identifying terms/items for qualitative research

Research Support Services Committee Approach: Referee’s Role Resolves style disagreements Manages interaction Brings survey researcher perspective Keeps an eye on source version Pushes for global decisions Liaison with research team for consultation and documentation

Research Support Services Committee Approach: Cognitive Interviews Allow to administer all or part of instrument Give a glimpse into thought processes Allow to see how different alternative terms work

Research Support Services Committee Approach: Focus Groups Permit to distinguish what is idiosyncratic Allow us to listen to how Rs. use language Allow to include more people in a shorter time Allow to include different national origin Rs and see if they reach consensus

Research Support Services Advantages of Committee/ Team Approaches Group process benefits Include different varieties of language in translation team Qualitative research that follows allows to incorporate the target population into process Relatively low cost Relatively quick

Research Support Services Selected Committee Translations: NSFG (U. Mich.) ATUS (BLS) NHSDA (RTI) ICARIS (Battelle) NMHS (Battelle) SLAITS Asthma Q. (Abt) Job Corps Student Q. (Battelle) Survey of Consumer Attitudes (U. Mich.) Catholic Voters Political Attitudes (BRS) McNair Program Evaluation (DIR) L.A. Latino Eye Study (USC) Women’s Health Initiative Q. (U. Mich.) WHO Health & Performance Q. (Harvard) PHDCN (Harvard) Project Bread (UMass-Boston) California Safe Schools (RAND) CAHPS Dialysis Center (RAND) National Children’s Study Pilot (Battelle) Survey of Bioterrorism Pre- paredness (N.Y. Academy of Medicine)

Research Support Services  Textual assessments  Translation appraisal  Holistic approaches, e.g., TRAPD, committee  Back translation  Pretesting with bilinguals, e.g.,  Splits  Double administrations  Debriefings  Probe interviews  Think alouds with Respondent or Translator  Focus groups with sample population Assessing Translations: some procedures (Harkness, Pennell, Schoua-Glusberg – ASA, 2003)

Research Support Services Backtranslation One person translates from source into target language. A second person translates the target language version back into the source language. A third person compares the original and the backtranslated source language versions. Discrepancies are investigated.

Research Support Services What is wrong with back- translation? It is a black box: we know what went in, we know what came out, yet we know nothing about the adequacy of the target language version.

Research Support Services Why give Q-by-Q specs to translators? In the absence of question-by-question specifications, translators make their own decisions…  to resolve ambiguities  to figure what “they” mean These are most often not documented or even explicit. They happen in the translator’s mind.

Research Support Services Research Conducted Goal: to examine translators’ decision processes Vehicle: Recorded a translators’ Committee meeting Method: Listened to tape searching for decisions made during discussions

Research Support Services Why this design? Committee discussion involves: –Review –Translation –Adjudication Makes decisions verbally explicit

Research Support Services Transcription Detail: 2 hours of committee meeting 11 pages of questionnaire 110 questions

Research Support Services Example 1 How would your parents feel if they found out you drank alcohol sometimes?  Not at all upset  A little upset  Pretty upset  Very upset

Research Support Services Discussion about “upset” Did “they” mean… Angry? Perturbed / Bothered? Given context, committee decided to translate upset as angry.

Research Support Services Example 2 I think sometimes it’s okay to cheat at school. Strongly Agree Agree Disagree Strongly Disagree

Research Support Services Discussion about ‘cheat’ Original Item: I think sometimes it’s okay to cheat at school. First translation: Creo que a veces está bien hacer trampa en la escuela. Referee: “¿Hacer trampa? Everyone agrees?” Orig. Tr.: To me it refers to copying in exams. Tr2:“You can also copy when you are doing homework or other schoolwork, that is cheating too, not only in exams.” “And you could cheat in sports, so we need to be specific. In this case this refers to school work, not to playing volleyball.” Resolution: Creo que a veces está bien copiar o hacer trampa en mis exámenes y tareas escolares. (I think sometimes it’s okay to copy or cheat in my exams and schoolwork.)

Research Support Services Example 3 How do you feel about your ability to care for your teen when they are sick or upset? Discussion about ‘upset’. In this case, the discussion centered around whether ‘upset’ meant angry, not feeling well, or bothered by something. The committee went with the latter option. (‘upset’: molesto)

Research Support Services Example 4 How do you feel about your ability to discipline your teen? Discussion about ‘discipline’. Do “they” mean: Punish? Teach them to behave properly? Set rules for them? Checked with client who opted for ‘punish’.

Research Support Services Example 5 How do you feel about your ability to obtain needed resources for your teen? Discussion about ‘resources’. Did “they” mean... Material resources? Assistance/help? Client asked to preserve the ambiguity as much as possible. Translated as ‘recursos’.

Research Support Services Example 6 During the past week, how often did you let this teen know you really care about him/her? Discussion about ‘care’. Did “they” mean… love? concern? Committee decide to include both: … how much you love him/her and are concerned about him/her

Research Support Services Providing Specifications: Ideal Model Three steps: 1. Deliver question-by-question specifications 2. Review original text with translation team member 3. Ask translators to make their decisions explicit and submit them for review.

Research Support Services Format for Documentation Q#Original Q.IssueD/Q Specify decision or question

Research Support Services Recent/Current Efforts to Standardize Translation Procedures U.S. Census Bureau translation guidelines International Social Survey Programme’s translation methods’ work European Social Survey Implemented Procedures in 20+ countries European Social Survey and Survey of Health, Aging and Retirement in Europe analyzing translation and documentation outputs CAHPS Cultural Comparability Task Force NCHS Translation Issues Forum CSDI (Comparative Survey Design and Implementation Translation Task Force)