Issues in Translation

1
Issues in Translation

• David L. Streiner, Ph.D., C.Psych.
• Professor, Departments of Psychiatry
• University of Toronto/McMaster University
• Maven-in-Residence, MCHRI

2
The Need for Translation

• Use the same scale in different countries
• Use the same scale in multicultural cities and
  countries
• CCHS 1.2 available in
• English
• French
• Punjabi
• Mandarin

3
The Goal of Translation

• To achieve equivalence between original and
  translated versions
• But, at least 19 different types of equivalence
  have been proposed
• Most agree there is a hierarchy of five or six
  key aspects

4
Levels of Equivalence

• Conceptual
• Item
• Semantic
• Operational
• Measurement

5
Equivalence

• Translation involved moving between (or even
  among) languages
• Also involves moving between cultures

6
Level 1 Conceptual Equivalence

• Do people in both cultures see the concept in the
  same way?
• This is a continuum, not a dichotomy

Concept doesnt exist in target culture
Same concept in both cultures
7
Conceptual Equivalence

• Happiness is not understood by poor Egyptian
  women
• Japanese concept of amae doesnt exist in West
  emotional dependence and more not captured in
  guilt or social embarrassment
• Concept of God very different in non-monotheistic
  religions (e.g., Buddhism, Hinduism, Shintoism)

8
Conceptual Equivalence

• Getting drunk is expected during festivals among
  Mapuche in Chile not evidence of alcohol abuse
• Similarly in Japan drunkenness in men expected
  in social situations
• Corporal punishment is part of normal child
  rearing in Chile not seen as abusive

9
Conceptual Equivalence

• Physical independence prized much more highly in
  North America and Europe than in Japan
• Sleepiness (side-effect of many meds) more
  tolerated in Japan than U.S.

10
Conceptual Equivalence

• Determining conceptual equivalence
• Review of ethnographic and anthropological
  literature
• In-depth interviews
• Focus groups
• Key informants

11
Level 2 - Item Equivalence

• Are specific items relevant and acceptable in
  target population?
• Asking about climbing stairs in settings of
  one-story buildings
• Asking about taboo topics (e.g., sex, income,
  negative feelings about elders)
• Questions may have to be reworded or replaced

12
Level 3 - Semantic Equivalence

• Looks at the meaning attached to each item
• I feel blue meaningless in many languages
• I can climb one flight of stairs and how many
  cigarettes smoked per day indices of wealth in
  rural China
• Terms from sports (e.g., get to first base)
  dont travel well

13
Semantic Equivalence

• A problem even within the same language
• We have really everything in common with
  America nowadays except, of course, language.
  Oscar Wilde
• The GHQ asks about needing a tonic
• In UK, a medicine that invigorates or strengthens
• In US and Canada, what you mix with gin

14
Semantic Equivalence

• In Spanish, comadre
• Godmother or best friend in Mexico, Texas
• The other woman in Nicaragua
• Chevy Nova didnt sell in South America
• No va Doesnt go
• Robo (Carwash) I steal
• Colgate Go hang yourself (Argentina)

15
Semantic Equivalence

• Date of childs birth 09/02/08
• February 8, 2009?
• February 9, 2008?
• September 2, 2008?
• Depends on the country

16
Level 4 Operational Equivalence

• Looks at the format, instructions, mode of
  administration
• Literacy level for self-administered scales
• Asking direct questions in some First Nations
  cultures
• Asking questions of elders

17
Operational Equivalence

• Visual analogue scales (VASs)

No Pain at All
Worst Pain Imaginable
18
Operational Equivalence

• Visual analogue scales (VASs)
• Person places an X corresponding to degree of pain

No Pain at All
Worst Pain Imaginable
X
No Pain at All
Worst Pain Imaginable
19
Visual Analogue Scales

• Seemingly, very simple task
• Actually, quite cognitively complex
• Must think of pain as a continuum
• Assess degree of pain
• Translate that to a distance
• Elderly and illiterate people have difficulty
  with this
• Had to change format

20
Visual Analogue Scales
Worst Pain Imaginable
No Pain at All
21
Likert Scales

• I am happy most of the time.
• Strongly Agree
• Agree
• Neither Agree nor Disagree
• Disagree
• Strongly Disagree

22
Likert Scales

• Have to deal with two concepts
• Agreement / Disagreement
• Strength of feeling
• May be difficult for cognitively impaired or low
  literacy people
• May be better to split each question

23
Likert Scales

• I am happy most of the time.
• Do you agree, disagree, or neither?
• If agree or disagree
• How strongly do you _____?
• Makes questionnaire longer, but easier

24
First Intermezzo

• Before the last stage (Measurement equivalence),
  we can start the translation process

25
Translation

• Step 1
• Two teams (2) people translate into target
  language
• Peoples first language should be target language
• Should know intent of scale as a whole and each
  item
• Allows them to go beyond strict semantic
  translation use idiomatic language
• Differences reconciled

26
Translation

• Step 2
• Again, two independent teams
• Have not seen original scale
• Do not know purpose of scale
• First language is source language familiar with
  idiomatic usage

27
Translation

• Step 3
• Committee looks at original and translated
  versions
• All members fluent in both languages
• Mix of people whose first language is source and
  target

28
Second Intermezzo

• Now that translation is done, and before the 5th
  stage of ensuring equivalence, make sure items
  work in target language
• Cannot use translation committee
• Because bilingual, may be brighter than average
  person
• Know both cultures
• Cognitive interviewing
• Should also be used in developing scales

29
Cognitive Interviewing

• Can take many forms
• Rephrasing
• Double interviewing
• Thinking aloud
• Probing

30
Rephrasing

• Respondent rephrases item, keeping meaning as
  close to original as possible
• Verbatim responses coded
• Fully correct
• Generally correct (one part altered or omitted)
• Partially wrong (but understood intent)
• Completely wrong

31
Rephrasing

• Problems also if respondent says
• I never heard of that.
• Could you please repeat that?
• or similar phrases
• Somewhat subjective

32
Double Interviewing

• Person first completes entire scale
• For 3-5 items, interviewer asks
• What led you to answer ___? or
• Tell me exactly how you came to that answer.

33
Thinking Aloud

• Similar to double interviewing
• Person does not fill out questionnaire first
• For each question, respondent thinks aloud how
  he/she chose the response

34
Probing

• After each question, person asked
• How easy or hard was it to answer?
• How sure are you of your answer?
• How did you feel about answering it?
• etc.

35
Which to Use

• If issue is understanding, better to use
  Rephrasing or Probing
• If person has to recall or estimate (e.g., How
  many times did you visit your GP last year?),
  may be better to use Thinking Aloud or Double
  Interviewing
• Dont have to be consistent for all items
• Sample to redundancy

36
Level 5 Measurement Equivalence

• Easiest to do
• Establish reliability and do some validity
  testing in target population
• More difficult to do
• Establish psychometric equivalence using
  confirmatory factor analysis (CFA) and/or item
  response theory (IRT)

37
Measurement Equivalence

• Three steps with CFA
• Do the items load on the same factors?
• If so, are the magnitudes of the factor loadings
  the same?
• If so, are the variances of the items the same?
• With IRT
• Look at differential item functioning comparing
  the original and translated versions
• Dont try this at home!

38
Translation

• Is it worth it?
• Same (or more) work than starting from scratch
• May still be differences
• On the plus side, end up with similar content and
  (ideally) factor structure

39
Finished

• Fini
• Fertig
• Acabado
• Geëindigdp
• Te?e??µ????
• Rifinito
• ?????