Académique Documents
Professionnel Documents
Culture Documents
RESEARCH METHODOLOGY
Abstract
Title. Translation of scales in cross-cultural research: issues and techniques
Aims. This paper is a report of a study designed to: (i) describe issues and techniques
of translation of standard measures for use in international research; (ii) identify a
user-friendly and valid translation method when researchers have limited resources
during translation procedure; and (iii) discuss translation issues using data from a
pilot study as an example.
Background. The process of translation is an important part of cross-cultural
studies. Cross-cultural researchers are often confronted by the need to translate
scales from one language to another and to do this with limited resources.
Method. The lessons learned from our experience in a pilot study are presented to
underline the importance of using appropriate translation procedures. The issues of
the back-translation method are discussed to identify strategies to ensure success
when translating measures.
Findings. A combined technique is an appropriate method to maintain the content
equivalences between the original and translated instruments in international
research. There are several possible combinations of translation techniques. However, there is no gold standard of translation techniques because the research
environment (e.g. accessibility and availability of bilingual people) and the research
questions are different.
Conclusions. It is important to use appropriate translation procedures and to
employ a combined translation technique based on the research environment and
questions.
Keywords: back-translation, bootstrapping statistics, instrument validation, International research, nursing, translation issues, translation techniques
Introduction
In cross-cultural studies, using previously developed instruments with good psychometric properties can save time and
effort. However, these instruments need to be culturally
386
Background
Issues related to translation
Cross-cultural researchers can use instruments that were
originally constructed in two or more languages. Or, they can
use measures that need to be translated as they were not
initially developed for cross-cultural research. Some issues of
translation overlap; others occur because of the type of
instrument. In other words, realistic problems in the translation procedure by cross-cultural researchers may differ
according to the nature of their studies and the characteristics
of instruments.
Translation techniques
Four strategies for maintaining content equivalence:
advantages and disadvantages
Importing a scale for use in another language or culture often
requires considerable effort by researchers to maintain the
quality of translation (Brislin 1970, Sechrest & Fay 1972,
Wang et al. 2006). Brislin (1970) offered four techniques for
387
maintaining the equivalence between the original and translated measures: (i) back-translation method; (ii) bilingual
technique; (iii) committee approach; and (iv) pretest procedure.
Back-translation is a well-known method to maintain
equivalence between the original and translated versions
(Behling & Law 2000); Brislins classic back-translation
model (1970) is widely used for instrument validation. Brislin
recommended an iterative process of repeated independent
translation and back-translation by a team of translators. This
approach requires several independent bilingual translators
(Triandis & Brislin 1984). A bilingual translator blindly
translates an instrument from the original language to the
target language; a second bilingual translator independently
back-translates the instrument from the target language to the
original language. Next, the two versions of the instrument
(original language and back-translated version) are compared
for concept equivalence. When an error is found in the backtranslated version, another translator attempts to retranslate
the item. This procedure continues until a team of bilingual
translators agrees that the two versions of the instruments are
identical and have no errors in meaning (Table 1).
The major weakness related to Brislins classic translation
model is that researchers cannot estimate how many independent bilingual translators are needed to get content
equivalence between the original and the translated versions;
accessibility and availability of qualified bilingual people who
have knowledge of the original (source) and target languages,
both cultures, and the focal area of the research is a key when
using this method. As an alternative, researchers often use a
team approach with two independent bilingual translators
(Yam et al. 2004, Thato et al. 2005); yet, this method may
amplify the problems of the translated measures when the
original and target languages have different structures, the
original instrument includes metaphorical or emotional
terms, and translation is conducted by unqualified bilingual
experts (Brislin 1970, Dunnigan et al. 1993, Russell & Sato
1995, Jones et al. 2001).
The use of the bilingual technique requires the administration of the instrument in both the original and target
languages to bilingual participants (Brislin 1970). The
responses of the participants to the two versions are
compared. When researchers identify a different response,
they need to consider the potential reasons for this discrepancy. Yet, bilingual participants responses may be different
from those of monolinguals (target population) because
bilingual people are acculturated to their host culture. This
acculturation places them in a separate population from the
monolingual population (McDermott & Palchanes 1992,
Sperber et al. 1994) and enable them to report different
responses even though the two versions of the measure
(original and target versions) have content equivalence (Lee
et al. 2002).
The committee approach uses a group of bilingual experts
to translate from the original to the target language (Brislin
1970). This approach provides a clearer version of the
translated instrument as one committee members mistake
can be more readily identified by the other committee
members (Brislin 1970). This method may be appropriate
to use when bilingual translators have a preference for a
target language and a limited number of persons are available
for the back-translation procedure. However, this method
requires more than three bilingual people and the accessibility
of bilingual people as translators is a key issue when applying
this approach.
The fourth method used for translation is the pretest
procedure (Brislin 1970). The pretest refers to a pilot study.
This procedure helps researchers identify potential problems
when the population, study conditions, and measures that
will be used for a larger study are the same as the pilot study.
Accurate information about the potential problem regarding
equivalence of translated measures can be obtained according
to the level of similarities of the study conditions. While the
back-translation method has frequently been used to maintain content equivalence in the translated version, there are
limitations when using Brislins classic back-translation
Table 1 Equivalent issues related to translation and decentering applied in the exemplar
Issues for maintaining content equivalence
Problems confronted
Applied solutions
Vocabulary equivalence
Idiomatic equivalence
Grammatical-syntactical equivalence
Experiential equivalence
Conceptual equivalence
388
The study
Aims
The aims of this study were to: (i) describe issues and
techniques of translation of standard measures for use in
international research; (ii) identify a user-friendly and valid
translation method when researchers have limited resources
during translation procedure; and (iii) discuss translation
issues using data from a pilot study as an example.
Instruments translated
We have chosen two instruments, the Sexual Abstinence
Efficacy Scale (SAES) and the modified Premarital Sexual
Attitude Scale (mPSAS), to illustrate the use of the combined
translation technique and the decentering applied in this
study. Both scales were translated from English to Korean in
order to use them in the pilot study identifying potential
predictors of sexual behaviour among Korean college
students. The combined translation technique included
the back-translation method, committee approach, and
pretest procedure using a monolingual sample. We applied
decentering to obtain vocabulary, idiomatic, grammatical
syntactical, experiential and conceptual equivalences between
English and Korean versions.
SAES
The SAES was used to assess sexual abstinence efficacy. The
tool, originally developed in Spanish and English, was designed
to measure sexual abstinence efficacy among high school and
college students (Norris et al. 2003). The scale consists of seven
items and one optional item. The optional item is only asked to
persons who have engaged in sexual intercourse, but who want
to refrain from sexual intercourse. Response options range
from 1 not at all sure to 4 extremely sure. Internal
consistency reliability using seven items ranged from an alpha
of 080 to 083 (Norris et al. 2003).
mPSAS
The original PSAS was used to assess premarital sexual attitude. The PSAS consists of 16 items measured on a 5-point
Likert scale (1 strongly disagree to 5 strongly agree)
assessing adolescent acceptance of sexual behaviour (e.g.
kissing, light and heavy petting, sexual intercourse) on four
levels of relationships (casual partner, steady partner, lover,
fiance/fiancee). The internal consistency reliability of the PSAS
was 089 (Treboux & Busch-Rossnagel 1995). For cross-cultural research using Korean college students, we modified the
PSAS. Sex worker/prostitute was included as the fifth level of
relationship assessing acceptance of each sexual behaviour
because Asian men are usually infected with STDs from a
commercial sex worker (Parish et al. 2003). Thus, the mPSAS
contains 20 items using a 5-point Likert scale.
389
Data analysis
Participants who completed the instruments were included in
the final data analyses (male 15, female 21; mean
age 2207 188; range 1824). We examined the
reliabilities of the translated measures using Cronbachs
alpha and bootstrap statistics which allows an examination
of stability of parameters with small samples. The bootstrapping method is a way to test the reliability of the data set with
pseudo-replicate datasets by re-sampling. This method
regards the researchers original sample as the population
and allows the researcher to obtain evaluated errors and to
evaluate estimated reliabilities (Opperdoes 1997). Cases from
the original data file are randomly selected with replacements
to generate other data sets (Bollen & Stine 1993). One
thousand bootstrap replications were conducted using SPSS
120 (SPSS Inc., Chicago, IL, USA) to estimate empirical
distributions of the internal consistency and provide the
ranges of Cronbachs alphas for the pilot study.
The translated instruments were used in a study designed
to assess the appropriateness of the measures for use in a
larger study on sexual behaviour of Korean college students.
Although researchers are working on the development of
effective sex education programmes for late adolescents,
these efforts are primarily focused on promoting consistent
condom use rather than sexual abstinence practice because of
their age and an existing high prevalence of sexual behaviour
among this population. However, a sexually abstinent person
not only includes a virgin but also a person who has
experienced sexual intercourse and subsequently decides to
become abstinent (Norris et al. 2003). Therefore, sexual
abstinence programme is not only applicable to virgins; the
permission to conduct the study from a university administrator in Seoul, Korea as there was no university ethics
committee. After that, the study was approved by the
University of Pittsburgh Institutional Review Board. Because
the study was anonymous, no written consent was necessary.
Findings
New combined translation technique
The SAES and mPSAS were translated from English to
Korean. Firstly, three independent bilingual people, including one of the researcher team, independently translated
the instruments. Each translated instrument was assessed by
two other bilingual people. Any differences identified
between the reviewed versions and their own translated
version were discussed in the committee meeting of the three
bilingual people. This procedure was continued until the
three translators agreed on the translated instruments
(Figure 2).
Secondly, a bilingual KoreanAmerican, who was an
undergraduate student, translated the Korean version measures back into English. Next, a monolingual English-speaking
person compared the original English version and the backtranslated English version. Additionally, the author of the
SAES served as a consultant during this process; she clarified
the meaning of items for translation and compared the two
English versions of the SAES. When the monolingual
reviewers identified a difference between the original and
the back-translated versions, they provided the translators
with detailed explanations of the usage differences in both
versions of the instruments. These differences and the
explanations were shared with the three bilingual translators
to retranslate the items. Based on the discussion, three
translators modified the wording of the instruments in the
Korean version. Then, the back-translator translated the
modified items from Korean to English again. The process
391
Table 2 Internal consistency in English and Korean versions of the instruments in the exemplar
English version
Modified Premarital
Attitude Scale (mPSAS)
Sexual Abstinence
Efficacy Scale (SAES)
SD ,
Korean version
Sample description
Alpha
Alpha
Bootstrap statistics
089
095
080083
083
Reliabilities
Both instruments had good internal consistency when used
with this sample of 36 Korean college students (see Table 1).
Cronbachs alpha for the mPSAS was 095 which was higher
than the original English version. The internal consistency
reliability of the SAES was consistent with the original
English version (a 083). The bootstrap method also
supported the good reliabilities and stabilities of the translated instruments (see Table 2).
Discussion
A combination of translation techniques is ideal for successful
translation (Brislin 1970, McDermott & Palchanes 1992,
Jones et al. 2001, John et al. 2006) and several possible
combinations of translation techniques exist. Researchers use
a combination technique without the back-translation
method, or they apply a single back-translation method to
obtain appropriate instruments for their studies. Thus, they
often report different psychometric properties such as internal consistency from the original studies using the measures.
The issue is whether true differences exist between the
original measures and the measures used in cross-cultural
research.
The strengths of using the combined techniques in the
current study were that a small number of bilingual people
was needed, and that the translation procedure was shorter
because a clearer translated version was obtained by using the
committee approach in the beginning. We obtained a more
reliable estimate of statistics from the dataset by performing
bootstraps. As a result of applying the new approach, we
successfully obtained translated content equivalent measures
with acceptable internal consistency reliabilities.
Although the translated instruments showed good reliabilities, several implications for future research were identified.
Firstly, a translated instrument needs to be revalidated with
the target population because of the potential distortions of
the items during the translation procedure (Streiner &
Norman 1989); however, we only examined the internal
consistencies of the instruments in the pilot study. There is
insufficient evidence to conclude that the translated measures
are valid instruments for assessing Korean college students
sexual behaviour. According to Paunonen and Ashton
(1998), instrument validation in cross-cultural research is
evaluated with the assessment of multiple psychometric
properties. Scale means, variance, reliabilities, criterion
validity and factorial validity are appropriate approaches
for instrument validation. The cultural differences and the
translation procedure often induce the failure of obtaining
invariance of the psychometric properties of the original and
the translated version of the instrument in cross-cultural
research. Adequate sample size, both within and between
cultures, is necessary to reduce sampling error. Thus, there is
a need to examine these five suggested psychometric properties of the translated scales with a larger sample for
instrument validation in cross-cultural research.
Secondly, high internal consistencies of instruments often
occur when there is item redundancy or an increasing number
of items (Streiner & Norman 1989, Nitko 2004). In the
mPSAS, the Cronbachs alpha may have increased with the
inclusion of additional items after the translation procedure.
Therefore, further investigation is needed to identify the
reasons for this problem (Streiner & Norman 1989). Itemtotal correlations and inter-item correlations with a larger
sample may provide an answer as to whether any items in the
Korean version of the scale are redundant (Sechrest & Fay
1972, Streiner & Norman 1989, DeVellis 2003). Lastly,
bootstrap statistics have a weakness when the original sample
is small and non-representative (Kline 1998). Thus, there is a
need to reexamine the internal consistencies with a larger,
more representative sample.
393
Conclusion
Cross-cultural research will help nurses to compare the findings
of previous studies and identify potential contributing factors to
develop effective nursing intervention programmes with the
goals of promoting health. Standardized translated measures,
which are crucial to conducting cross-cultural research, can be
obtained when researchers are attentive to the translation
procedure. However, there is no gold standard for translation
techniques because the research environment (e.g. accessibility
and availability of bilingual people) and the research questions
are different. Our findings give insight into what an appropriate
translation approach is when researchers have limited sources.
Acknowledgements
The authors would like to acknowledge Dr A.E. Norris,
Associate Professor, Boston College School of Nursing for her
guidance about back-translation method. The authors thank
to Drs Thelma E. Patrick and Susan M. Sereika for their
reviews of a preliminary paper. The authors are grateful to
Drs Betty Braxter and Scott Weber, University of Pittsburgh
School of Nursing for their supports and helpful guidance in
relation to the back-translation method. The authors thank to
the translators.
Author contributions
ESC, KK and JE were responsible for the study conception
and design and the drafting of the manuscript. ESC per394
References
Behling O. & Law K.S. (2000) Translating Questionnaires and other
Research Instruments: Problems and Solutions. Sage Publications,
Inc., Thousand Oaks, CA, USA.
Bollen K.A. & Stine R.A. (1993) Bootstrapping goodness-of-fit
measures in structural equation models. In Testing Structural
Equation Models (Bollen K.A. & Scott Long J., eds), Sage, Newbury Park, CA, USA, pp. 111135.
Brislin R.W. (1970) Back-translation for cross-cultural research.
Journal of Cross-Cultural Psychology 1, 185216.
Brislin R.W. (1976) Comparative research methodology: crosscultural studies. International Journal of Psychology 11, 215229.
Brislin R.W., Lonner W.J. & Thorndike R.M. (1973) Cross-Cultural
Research Method. Wiley, NY, USA.
DeVellis R.F. (2003) Scale Development: Theory and Applications.
Sage, Thousand Oaks, CA, USA.
Dunnigan T., McNall M. & Mortimer J.T. (1993) The problem of
metaphorical nonequivalence in cross-cultural survey research:
comparing the mental health statuses of Hmong refugee and general population adolescents. Journal of Cross-Cultural Psychology
24, 344365.
John M.T., Hirsch C., Reiber T. & Dworkin S.F. (2006) Translating
the research diagnostic criteria for temporomandibular disorders
into German: evaluation of content and process. Journal of
Orofacial Pain 20, 4352.
Jones P.S., Lee J.W., Pillips L.R., Zhang X.E. & Jaceldo K.B. (2001)
An adaptation of Brislins translation model for cross-cultural
research. Nursing Research 50, 300304.
Kline R.B. (1998) Principles and Practice of Structural Equation
Modeling. Guilford, NY, USA.
Lee J.W., Jones P.S., Mineyama Y. & Zhang X.E. (2002) Cultural
differences in responses to a likert scale. Research in Nursing &
Health 25, 295306.
McDermott M.A. & Palchanes K. (1992) A process for translating and
testing a quantitative measure for cross-cultural nursing research.
Journal of the New York State Nurses Association 23, 1215.
Nitko A.J. (2004) Educational Assessment of Students, 4th edn.
Prentice Hall, Englewood Cliffs, NJ, USA.
Norris A.E., Clark L.F. & Magnus S. (2003) Sexual abstinence and
the sexual abstinence behavior scale. Journal of Pediatric Health
Care 17, 140144.
Opperdoes, F. (1997). Bootstrapping. Retrieved from http://
www.icp.ucl.ac.be/opperd/private/bootstrap.html on 8 July
2005.
Parish W.L., Laumann E.O., Cohen M.S., Pan S., Zheng H., Hoffmann I., Wang T. & Hang K. (2003) Population-based study of
Chlamydial infection in China. Journal of the American Medical
Association 289, 12651273.
Paunonen S.V. & Ashton M.C. (1998) The structured assessment of
personality across cultures. Journal of Cross-Cultural Psychology
29, 150170.
395