Collaborative and Iterative Translation:

An Alternative Approach to Back

Translation
ABSTRACT

Susan P. Douglas and

C. Samuel Craig

Submitted April 2006

Accepted September 2006
Journal of International Marketing
2007, American Marketing Association
Vol. 15, No. 1, 2007, pp. 3043
ISSN 1069-031X (print)
1547-7215 (electronic)

30

Back translation is the technique most commonly used to check

the accuracy of translation in survey research. Although it
results in a direct or literal translation, it does not address
issues of conceptual equivalence. Furthermore, if it is not combined with pretesting, it does not address issues of comprehension and meaning to the respondent. This article discusses
some of the problems of relying exclusively on back translation
in isolation. It suggests the use of a collaborative, iterative
approach for translating questionnaires for international marketing research.
As companies expand further into international markets,
penetrating emerging markets, such as China, India, and
Brazil, as well as markets within the former Soviet Bloc, the
Middle East, and Latin America, the conduct of international
marketing research is becoming increasingly important and
more complex. Not only are researchers increasingly using
cross-national surveys to investigate customers attitudes
and interests, but advertising and product concepts are also
being tested on a global basis. The diversity of languages spoken in these countries means that the effective translation of
questionnaires and instruments used in experiments is a
central concern. To obtain meaningful results, it is essential
to establish equivalence of meaning in each language and to
ensure that each respondent and interviewer clearly understands the questionnaire and instructions.
A procedure that is commonly used to test the accuracy of
translation in multicountry research is back translation (Brislin 1970, 1980). In this procedure, a bilingual native of the
target country translates a questionnaire into the target language.1 A bilingual native speaker of the source language
then translates it back into the source language. The original
and back-translated versions are then compared for differences and comparability. The accuracy of the back-translated
version is considered an indicator of the accuracy of the target translation. Back translation was initially developed for
situations in which a researcher was not familiar with the
target language but wanted some assurance that respondents
were indeed being asked the same question in that language
(Harkness 2003). The procedure was believed to provide
insights into potential errors when no other means were
available to assess the accuracy of the translation.

Back translation is the most commonly used approach in

marketing to help identify problems and egregious errors in
translation. However, it does not necessarily ensure equivalence in meaning and concepts in each country, and thus the
translation that is adapted to the culture of the target language might not be accurate. This is particularly likely to be
an issue when idioms or colloquial language is used, as is
often the case in marketing surveys. The purpose of this article is to describe the use of back translation and the problems
associated with reliance on back translation in isolation. We
propose an alternative, collaborative, iterative approach to
translation that should result in a translated instrument that
more accurately reflects the meaning of the source language
in the target culture. First, we review commonly used translation practices and identify their limitations. Second, we
discuss the steps needed to ensure appropriate translation,
depending on the specific situation. Finally, we recommend
a five-step iterative procedure that can lead to improved
translations.
Historically, in the social sciences, back translation has been
widely used to test the accuracy of the translation and to
detect errors in translation (Brislin 1970, 1980). However, in
recent years, the use of parallel or double translation (i.e.,
two translated versions) has been advocated as a preferred
method of achieving equivalence in meaning (Hambleton
1993, 1994; Van de Vijver and Hambleton 1996). This can be
used in conjunction with a committee approach to review
the meaning and equivalence of translations and to select the
best translation (Harkness 2003; Harkness and SchouaGlusberg 1998). Extensive checking, pretesting of the translation, and debriefing are essential to ensure a reliable and
accurate translation.

RELIANCE ON BACK
TRANSLATION

In marketing, back translation is still the primary method

used to check translation accuracy. A review of Journal of
International Marketing since its inception (19932005; i.e.,
52 issues) identified a total of 45 articles that report surveys
that use multiple languages. Of these surveys, 17 were in
European languages, 3 were in Russian and Kazakh, 1 was in
Arabic, and 28 were in Asian languages (Japanese, Thai, Chinese, Vietnamese, and Korean). In 75% (34) of these, back
translation was used, and in 6 cases, a committee checked
items in the back translations for equivalent meaning across
languages. In 5 cases, parallel or double translation was
undertaken, and a committee reviewed and adjudicated
these to determine which translation should be used. In 1
case, a computer system (IADS) capable of handling multiple
languages was used, and in 3 cases, no back translation was
undertaken. Only 14 cases used pretesting of the questionnaire to ensure that the respondents clearly understood the
questions. This typically entailed asking respondents the

Collaborative and Iterative Translation

31

meaning of the question and further probing their understanding of the meaning or significance of the questions.
The review suggested that the most common form of checking the accuracy of translation in marketing studies was back
translation. However, although there were limited details in
many cases, it appears that beyond some modification after a
single translation or back translation iteration by bilingual
translators, relatively few additional checks or tests were
undertaken to ensure the adequacy of the translation. A
mechanism that is particularly useful for detecting translation problems, pretesting, was used in fewer than one-third
of the studies.
As we noted previously, although back translation can potentially identify translation errors, in isolation, it is subject to
several limitations (Brislin 1970; Van de Vijver and Hambleton 1996). Because bilingual translators are fluent in both the
relevant languages, they do not necessarily use a language in
the same way as monolingual people and may have common
ways and conventions for moving from one language to
another. Bilingual translators may also be able to make sense
of a poorly written target translation (Brislin 1970). Because
back translation provides a direct or literal translation from
one language to another, it is possible to move from one language to another and back again without capturing the
intended sense of the statement. For example, this may occur
with idioms. Harkness (2003) notes an example from a German general social survey. An item in the survey is Das
Leben in vollen Zgen geniessen. Literally translated into
English, this means Enjoy life in full trains. (One reading
of Zgen is the plural of train; the other, idiomatic meaning is in full draughts.) A back translation may produce the
same wording as the original German, which would suggest
that the translation was accurate. However, the more appropriate translation into English would be Live life to the full
or, in American English, Live life to the fullest.
The limitations of translating an existing questionnaire by
relying solely on back translation can be observed in the
work of Douglas and Nijssen (2003). In research conducted
in the Netherlands, they used the CETSCALE (Shimp and
Sharma 1987), which was initially developed in the United
States. They tested two versions of the CETSCALE, a literal
back-translated version and a modified version that
attempted to address differences in consumer ethnocentrism
in the Netherlands, namely, the absence of domestic brands
in certain product categories. Consequently, an item such as
It is not right to buy foreign products had limited relevance. Because of this problem, the authors conducted indepth interviews to examine the meaning and significance of
the construct to Dutch consumers. They also obtained reac-

32

Susan P. Douglas and C. Samuel Craig

tions to the Dutch version of each back-translated question.

On the basis of qualitative research, they developed a modified version of the questionnaire that they believed to be
more faithful to the construct as it existed in the Netherlands. Although the back translation did not change the
meaning of any of the original scale items, the modified
questions were clearer and better adapted to the Dutch concept of consumer ethnocentrism. The analytical portion of
the study found that the modified questionnaire was also
superior.
In addition, back translation assumes an etic approach to linguistic translation or, in other words, that there is always an
equivalent word or construct in the target language. As a
result, in some cases, a complete phrase may be necessary to
clarify a construct. For example, Triandis (1972) states that
the concept of philotomo is unique to the Greek culture and
may need to be translated as a phrase such as the love of
honor to capture the full meaning. Consequently, back
translation is likely to be most useful when a literal or direct
translation is required, but it is less helpful when idioms
need to be translated or when the equivalence of a term or
construct in another language needs to be established.
The problems associated with back translation suggest that
simply relying on the back translation of a questionnaire
without a careful review and an equivalence check is
unlikely to provide an appropriate solution to determining
whether a translation is adequate. Subtle nuances in the use
of language, idioms, and the lack of equivalence of words or
phrases also need to be considered. This implies that direct
or literal translation is rarely adequate in and of itself.
Rather, some adjustment in the use of language structure or
phraseology is typically necessary to ensure equivalence in
meaning (Harkness 2003).
The complexity of translation, the subjective character of
translation assessment, and the multiple skills required (i.e.,
both linguistic ability and skills in questionnaire development) imply that a team-based approach to translation is
needed. In the social sciences, this has been found to provide
the richest output in terms of alternative translations and a
balanced critique of alternative versions (Harkness and
Schoua-Glusberg 1998; McKay et al. 1996). A team can bring
together the mix of required skills and the disciplinary
expertise required. Team members need to have knowledge
of the study and of questionnaire design, as well as the cultural and linguistic skills to translate the questionnaire into
appropriate versions of the target language when necessary.

COLLABORATIVE APPROACHES
TO INSTRUMENT
TRANSLATION

Two main forms of collaborative-based work can be used in

questionnaire translation: the committee approach and the

Collaborative and Iterative Translation

33

expert team approach. In the committee approach, much of

the work is done together, with collaborators working in a
group. In expert team approaches, team members work individually rather than as a group. This procedure may be
appropriate, for example, when the team is geographically
scattered and operates as a virtual team.
Regardless of the approach taken in developing a final translation of a questionnaire, five basic stages are involved:
translation, review, adjudication, pretesting, and documentation (Harkness 2003). Documenting translation procedures is
particularly desirable when the questionnaire is likely to be
repeated (e.g., an annual or biannual survey). As we noted
previously, these stages often result in an iterative process.
Review or adjudication may result in retranslation, and
pretesting may result in a further round of translation review
and testing.
A committee-based application of this procedure begins with
a form of translation, commonly, parallel translation
(Schoua-Glusberg 1992). In this case, two or more translators
make independent, parallel translations of the questionnaire
into the target language (Brislin 1980). A review meeting is
then held in which the translators and at least one independent reviewer discuss the various versions and decide on a
version for final review. An adjudicator may be present at the
meeting and be a reviewer, or the final review version may be
passed on for adjudication. In some cases, the researchers
conducting the study may act as adjudicators, especially if
they are familiar with the appropriate languages.
In many cases, committee-based review and adjudication are
merged, depending on team expertise, schedules, previous
experience with similar surveys or languages, and so forth.
Alternatively, two committee rounds can be used, one to
review and revise the translated versions and one to decide
whether to accept the revised version. This may be appropriate if the researchers need time to consult on which version
to adoptfor example, if the questionnaire is complicated or
highly technical. This may also be desirable if the
researchers are not familiar with one or more of the languages involved.

NEW VERSUS EXISTING

RESEARCH INSTRUMENTS

34

In approaching multicountry research translation,

researchers may encounter two different situations. The first
situation occurs when a new instrument is being developed
in multiple languages, and the second situation occurs when
an instrument has already been developed in one language
and needs to be translated into another language or multiple
languages.

Susan P. Douglas and C. Samuel Craig

In the first case, there is an opportunity to eliminate the dominance of the structure of a source language or culture and to
establish equivalence of different questions in each language
(Werner and Campbell 1970). When the content of the questionnaire has been established, a team of bilingual or multilingual researchers and translators, preferably a researcher
from each country and/or linguistic context, should examine
it. A researcher establishes a version in his or her own language and then translates it, if necessary with a translator,
into an agreed-on common language. The team then reviews,
discusses, and harmonizes each version until a comparable
and, as far as possible, equivalent version is established in
each language (Van de Vijver and Leung 1997). In some
cases, when the same concept cannot be expressed or tapped
in the same way in each language, this may result in emic
questions in a version. This is likely to occur with attitudinal
scales or with activities such as sports, which vary in popularity from one country to another.
In the second, more common case, in which an instrument
has already been established in one language, questions need
to be translated so that they capture the same meaning in
each language. As we noted previously, this is complex and
typically requires an iterative process and multiple revisions
of the translation to ensure that an appropriate translation is
achieved, particularly when multiple languages are
involved. In the remainder of the article, we focus primarily
on this case, which is the situation that most international
marketing researchers face.
Most of the literature related to translation and, in particular,
to the use of back translation has focused on translation of
the instrument in general. However, there are important
issues related to the type of data being collected. International surveys typically involve one or more of three different types of data: socioeconomic, behavioral, and attitudinal
data. Socioeconomic data are usually the easiest to translate
because similar or equivalent categories can typically be
identified. Behavioral categories require the identification of
similar types of behavior that are equivalent. This may be
more problematic, especially when the relevant behavior is
cognitive rather than physical. Most difficulties are likely to
occur with attitudinal data because the construct may differ
from one language to another. Consequently, it is not just linguistic equivalence that needs to be considered but also
other equivalence issues, such as category equivalence, functional equivalence, and construct equivalence (see Craig and
Douglas 2005).

INDIVIDUAL ITEM ISSUES

Types of Questions

The relevance of each of these issues depends to a substantial degree on the nature of the question. Sociodemographic
questions are most likely to raise the issue of category equiv-

Collaborative and Iterative Translation

35

alence related to employment, position with the firm, marital

status, and so forth. Here, it is important to ensure that categories are equivalent, have similar status, and perform similar functions in each context under examination. For example, the tasks performed by a lawyer in the United States
would be performed by either a barrister or a solicitor in the
United Kingdom. Racial or regional demarcations also may
be somewhat ambiguous, and in some cases, there may be
reluctance to answer such questions.
Questions related to lifestyle and behavior are the most
likely to raise the issue of functional equivalence. For example, questions related to washing or doing the laundry would
need to consider the mode and circumstances of washing
for example, whether hot water is available, whether washing machines are commonly used, or whether washing is
commonly done by domestic help or by services outside the
home. Similarly, engaging in sports may be viewed either as
a purely social activity or as a means of gaining exercise and
improving health. Therefore, questions about leisure-time
activities would need to take this into consideration.
Attitudinal or abstract questions are the most likely to raise
the issue of conceptual equivalence. The linguistic equivalence of words such as happiness, grief, or mourning
may have different connotations and tap other related constructs (Smith 2003). Equally, an apparently equivalent word
may reflect different levels of intensity. Use of multiple indicators that are linguistically distinct may help resolve this
problem (Przeworski and Teune 1966) but only if different
terms do not replicate the same differences in intensity.
Individual words can also present difficulties in obtaining an
accurate translation. Soriano and Foxall (2002) apply Mehrabian and Russells emotionality scales in Venezuela. They
employ what can best be described as an iterative backtranslation approach with pretesting to arrive at the final
instrument. Of the 36 adjectives used in the English version
of the scale, 20 were translated exactly. Either the remaining
16 adjectives did not have a direct Spanish equivalent, or the
preferred usage varied between Spain and Venezuela. For
example, the English word annoyed could be translated as
enfadado or molesto. The preferred choice in Spain would
be enfadado, but in Venezuela, it should be translated as
molesto. Some word pairs had no direct equivalent in Spanish, and a different pair needed to be used. Notably, most of
the translation difficulties were detected through pretesting
and not through back translation, which accurately reproduced all but three of the adjective pairs used in the scale.

Scale Anchoring

36

In addition to considering the translation of questions in a

survey, establishing equivalency in translating response

Susan P. Douglas and C. Samuel Craig

scales and response format is also important Specifically, it

is necessary to ensure that endpoints and intervals on a scale
are equivalent (Mullen 1995). In this context, three alternatives have been suggested to improve equivalence and to
mitigate language-related issues. These include using nonverbal scales, limiting the number of response options, and
better calibrating label strength.
First, the use of nonverbal scales, such as visual or numerical
scales, can aid in avoiding some of the difficulties in establishing linguistic equivalence of response scales. However,
they do not completely eliminate translation issues, because
the instructions for using the scale need to be translated as
well as the meaning of the scale and the endpoints of the
scales. Nonverbal and numeric scales are also subject to their
own problems of differences in meaning and equivalence
across cultures. Numeric meaning of endpoints of scales may
vary from country to country. For example, in Germany, 1
typically refers to the first or best, resulting in a reversal of
numeric scales compared with the order that is typically
implied in the United States. Consequently, they may not
necessarily be equivalent.
A second possible solution is to use dichotomous scales with
responses, such as yes/no, favor/disapprove, or agree/disagree, which have equivalents across languages. However,
this assumes the feasibility of writing a question with a
dichotomous response, which may not necessarily be easy
with attitudinal questions. It also requires the establishment
of an equivalent no-opinion or indifferent category and
also a do-not-know category.
The third alternative is to calibrate the response scale so that
the strength of verbal scales used is equivalent across languages. Here, two cases may be identified. In the first case,
only the ends of the rating or ranking scale are labeled. This
typically poses fewer problems than when each position on
the scale needs to be translated. However, there may be difficulties in finding translations of points such as strongly
agree/strongly disagree and somewhat agree/somewhat
disagree that capture the same level of intensity across languages. For example, in the Euro barometer, which measures
social and political attitudes in the European Union on an
annual basis, the French and English scales differ in structure and, to a lesser extent, in semantics (Harkness 2003).
Both use the semantic dimension of agree, but in the French
scale, the use of daccord/pas daccord suggests a unipolar
scale, whereas the English scale uses a bipolar construction
in which the wording is linguistically symmetrical with the
endpoints modified by strongly. Equally, the do-notknow category is cannot choose in English compared
with ne sais pas or do not know in French.

Collaborative and Iterative Translation

37

Even if the words are accurately translated, there still may be

issues of equivalence in meaning. Reliance on back translation will provide accurate translations but may fail to provide words that generate equivalent responses across countries. As we indicated previously, words in isolation pose
greater difficulty because they lack a context to help the
translator render an accurate translation. Voss and colleagues
(1996) examine the semantic comparability of adjectives
used to label points of scales in three languages (English,
Japanese, and Chinese). They employ two techniques,
numeric estimation and line production, to assess interword
distances between adjectives. Both techniques produced
comparable results. This approach enables the researcher to
calibrate the meaning more precisely and to develop scales
with equivalent scale points across countries. In addition to
using a formal methodology to achieve metric equivalence,
Voss and colleagues use meetings and discussions among the
translators to arrive at comparable words.

Pretesting the Questionnaire

After the final version of the translation has been approved,

the next stage is to pretest the translation for comprehension,
clarity, and coverage. Here, it is important to ensure not only
that the translation represents a faithful rendering of the original text but also that the relevant target population clearly
understands its meaning. Attention should be paid to everything included in the study, such as interviewer instructions,
show cards, and other visual aids. Various procedures can be
used to evaluate a translated questionnaire, including both
qualitative and quantitative approaches, though qualitative
procedures are more common because they provide richer
insights into the exact nature of translation issues.
Qualitative tests include, for example, feedback (debriefing)
from monolingual or bilingual respondents and field staff. In
this case, the questionnaire is administered to either monolingual or bilingual respondents, who are then asked to provide feedback about their understanding of the meaning of
questions, the ease of comprehension, clarity, and so forth.
Typically, it is desirable to conduct pretests with both monolingual and bilingual respondents because, as we noted previously, bilingual peoples use of language, vocabulary, and
interpretation of terms may differ. In the case of bilingual
respondents, half are given the question in one language, half
are given the question in another language, and the similarity
of responses is compared. Field staff who have administered
a pretest can provide feedback on the clarity of questions,
difficulties in respondents comprehension, and so forth.
This may be particularly desirable with semiliterate or lowincome respondents because they may not understand certain words, concepts, or terms. Local terms or usage may also
be common among such populations. Similarly, the ability to
understand and use support materials, such as show cards

38

Susan P. Douglas and C. Samuel Craig

and response scales, can be evaluated. Focus groups can also

be used to evaluate questionnaires and to probe the meaning
of different words and phrases. This may be especially helpful in the case of abstract attitudinal questions.
Quantitative methods can also be used to evaluate pretests
when a sufficiently large sample is available. These include
tests that are based on item response theory, double ballot
administration, and bilingual fielding splits (Harkness and
Schoua-Glusberg 1998). In the case of double ballot administration, bilingual respondents are provided with both language versions of the questionnaire, and differences in
responses are compared. The results of pretesting may result
in some modification or readjustment of the questionnaire,
in which case it could be pretested again. As we noted previously, this becomes an iterative process until a satisfactory
version of the questionnaire is developed.
In international marketing research, translation is a critical
issue in the development of a valid and effective data collection instrument. Although the starting point for any research
project is the underlying issue or conceptual framework
under investigation, the vehicle that drives survey research
is the questionnaire. International marketing research necessarily involves different countries in which different languages or language variants are spoken. It is critical to ensure
that translated questions have equivalent meaning in each
context, are devoid of cultural bias, and can be clearly understood by all respondents.

RECOMMENDED APPROACH
TO QUESTIONNAIRE
TRANSLATION

For the reasons we discussed previously, a single back translation in isolation does not provide a reliable instrument in
the target language. Although this helps ensure a correct literal translation, it does not help assess whether the questions
have equivalent meaning or are devoid of cultural bias. Even
in catching errors, much depends on the back translators
understanding of a question and its purpose. Furthermore, a
mechanistic application of back translation, without adjustments, is unlikely to result in an instrument that will produce reliable and valid results. Rather than continue to
employ back translation, we recommend that researchers follow the steps we outline subsequently when translating
questionnaires into other languages. Following these steps
will result in a questionnaire that produces reliable and valid
results. We summarize the steps in Figure 1 and elaborate on
them in the following subsections.
The limitations of back translation mean that it is critical to
use a team or committee approach to ensure a reliable and
effective translation. This team or committee should consist
of people who are familiar with each of the cultures in which
research is being conducted and the different languages

Collaborative and Iterative Translation

Employ a Collaborative
Approach

39

Figure 1.
Collaborative and Iterative
Questionnaire Translation

used. The collaborative approach ensures that different

points of view are represented. Furthermore, it is iterative by
nature and helps ensure that the best translation evolves.
When a new set of questionnaires is being developed for a
project that will be conducted in multiple languages and different cultural contexts, it is important to ensure not only
that the questions have equivalent meaning in each context
but also that the questionnaire is decentered, which often
requires multiple iterations of translation. This helps ensure
that one culture or language structure does not dominate
other questionnaire versions. When research has already
been conducted in one country and a questionnaire needs to
be translated into another language or multiple other languages, the committee needs to ensure not only that the target language version is faithful to the source version but also
that it will elicit equivalent types of responses.

Establish Equivalence

40

Different types of data (i.e., questions related to socioeconomic, behavioral, and attitudinal data) all pose different
issues and thus need to be examined to ensure that they are
couched in terms that are similar in their context to the
source questions and will elicit equivalent responses. Category equivalence and functional equivalence are relatively
straightforward to achieve, but they still require knowledge
of the local conventions and the cultural context. Construct
equivalence is more difficult to achieve because there are

Susan P. Douglas and C. Samuel Craig

instances in which the construct under examination is

unique to a particular culture. A collaborative approach with
researchers from each culture can help identify and resolve
issues of construct equivalence. Similarly, scale anchors,
visual stimuli, and interviewer instructions need to be translated so that they are understood and elicit equivalent
responses in each language. Again, some degree of cultural
adaptation may be required.
After the equivalence of the questions has been examined to
ensure that they are expressed in similar terms and are likely
to elicit similar responses, the next step is to conduct an initial translation. When there is an existing questionnaire,
there should be independent, parallel translation into the
target language. Two separate translators should each translate the questionnaire into the target language. A review
meeting should then be held with the translators and an
independent reviewer to decide on the final version. A further round may be needed to resolve inconsistencies and
ensure that the questionnaire accurately captures the same
meaning in each country.
After the researchers and translators arrive at what they
believe is an appropriate instrument, the next step is to
pretest each language version of the questionnaire. This is an
important step and one that is often ignored. Only 31% (14 of
45) of the studies reported in Journal of International Marketing incorporated this step. The initial version of the questionnaire can be pretested on monolingual or bilingual
respondents. Pretesting, followed by debriefing, can find
problems of comprehension or meaning that were not identified previously. Issues identified in pretesting are referred
back to the committee to be incorporated into the final version of the questionnaire. Without this step, there is no way
to know how actual respondents (versus translators) will
react to the instrument.
In addition to employing a collaborative approach and establishing equivalence, it is important for the researcher not to
view the initial translation as the final translation. There is a
natural tendency to want to get the research into the field
and to get the results as quickly as possible. However, without an accurate translation of the instrument, the results of
the research may not be valid. Committee and team
approaches are intrinsically iterative because there will
inevitably be differences of opinion and interpretation that
need to be resolved. Multiple iterations, though time consuming, help ensure that the translated instrument is appropriate for the new context.

Conduct an Initial Translation

Pretest the Questionnaire

Adopt an Iterative Approach

Although this approach may appear somewhat complex and

time consuming, it is nonetheless essential to consider the

Collaborative and Iterative Translation

41

issues involved at each step to ensure that a reliable and conceptually equivalent translation is achieved. The time taken
in this approach will depend to a large degree on the experience of the researcher or company that is undertaking the
approach and the number of languages involved. When a
market research company or academic researcher conducts a
survey in a country or language for the first time, it is important to undertake this approach to establish equivalence and
to ensure the integrity of the translation. A local research
company, a local manager, or an academic researcher familiar with the language can be used at the committee stage to
check the translation, help select the final version, and
decide whether revisions are needed after the pretesting.
Conversely, if the company or academic researchers already
have extensive experience in conducting similar surveys in a
given country or language, they are likely to be already familiar with relevant linguistic equivalents; thus, parallel translation may not be necessary, and the checking process is likely
to be more expeditious. However, pretesting is essential to
detect any potential errors, though fewer modifications and
only one iteration are likely to be needed.

CONCLUSION

Instrument translation is a crucial element when conducting

international marketing research. Unless careful attention is
paid to equivalence in meaning and to the constructs tapped
in each context, the results will not be comparable from one
linguistic context to another. As a result, inappropriate conclusions may be drawn, thus negating the value of the
research. Reliance on back translation as the sole means of
achieving an accurate foreign language translation is
unlikely to result in equivalence. Consequently, researchers
need to employ the proposed iterative approach to ensure
that valid and reliable multicountry instruments are
developed.

NOTE

1. Target language is the language into which the translation

is being made.

REFERENCES

Brislin, Richard W. (1970), Back Translation for Cross-Cultural

Research, Journal of Cross-Cultural Psychology, 1 (3), 185216.
(1980), Translation and Content Analysis of Oral and Written Material, in Handbook of Cross-Cultural Psychology, Vol. 1,
Harry C. Triandis and J.W. Berry, eds. Boston: Allyn & Bacon,
389444.
Craig, C. Samuel and Susan P. Douglas (2005), International Marketing Research, 3d ed. Chichester, UK: John Wiley & Sons.
Douglas, Susan P. and Edwin J. Nijssen (2003), On the Use of Borrowed Scales in Cross-National Research: A Cautionary Note,
International Marketing Review, 20 (6), 62142.
Hambleton, R.K. (1993), Translating Achievement Tests for Use in
Cross-National Studies, European Journal of Psychological
Assessment, 9 (1), 5765.

42

Susan P. Douglas and C. Samuel Craig

 (1994), Guidelines for Adapting Educational and Psychological Tests: A Progress Report, European Journal of Psychological Assessment, 10 (3), 22944.
Harkness, Janet A. (2003), Questionnaire Translation, in CrossCultural Survey Methods, Janet A. Harkness, Fons J.R. van de
Vijver, and Peter Ph. Mohler, eds. New York: John Wiley & Sons,
3556.
and A. Schoua-Glusberg (1998), Questionnaires in Translation, in ZUMA-Nachtrichten Spezial, No. 3: Cross-Cultural Survey Equivalence, Janet A. Harkness, ed. Mannheim, Germany:
ZUMA.
McKay, R.B., M.J. Breslow, R.L. Sangster, S.M. Gabbard, R.W.
Reynolds, J.M. Nakamoto, and J. Tarnai (1996), Translating Survey Questionnaires: Lessons Learned, New Directions for
Evaluation, 70, 93105.
Mullen, Michael R. (1995), Diagnosing Measurement Equivalence
in Cross-National Research, Journal of International Business
Studies, 26 (3), 57396.
Przeworski, A. and H. Teune (1966), Equivalence in CrossNational Research, Public Opinion Quarterly, 30 (Winter), 551
68.

THE AUTHORS
Susan P. Douglas is PaganelliBull Professor of Marketing and
International Business (e-mail:
sdouglas@stern.nyu.edu), and
C. Samuel Craig is Catherine
and Peter Kellner Professor of
Marketing (e-mail: scraig@stern.
nyu.edu), Stern School of
Business, New York University.

ACKNOWLEDGMENTS
The authors acknowledge helpful
comments from Dawn Lesh, Durairaj
Maheswaran, and Tom Meyvis on a
previous draft of this article.

Schoua-Glusberg, A. (1992), Report on the Translation of the Questionnaire for the National Treatment Improvement Evaluation
Study. Chicago: National Opinion Research Center.
Shimp, T.A. and S. Sharma (1987), Consumer Ethnocentricism:
Construction and Validation of the CETSCALE, Journal of Marketing Research, 24 (August), 28089.
Smith, Tom W. (2003), Developing Comparable Questions in
Cross-National Surveys, in Cross-Cultural Survey Methods,
Janet A. Harkness, Fons J.R. van de Vijver, and Peter Ph. Mohler,
eds. New York: John Wiley & Sons, 6992.
Soriano, Mirella Yani and Gordon R. Foxall (2002), A Spanish
Translation of Mehrabian and Russells Emotionality Scales for
Environmental Consumer Psychology, Journal of Consumer
Behavior, 2 (1), 2336.
Triandis, Harry C. (1972), The Analysis of Subjective Culture. New
York: John Wiley & Sons.
Van de Vijver, Fons and R.K. Hambleton (1996), Translating Tests:
Some Practical Guidelines, European Psychologist, 1 (4), 8999.
and Kwok Leung (1997), Methods and Data Analysis for
Cross-Cultural Research. Thousand Oaks, CA: Sage Publications.
Van Herk, Hester (2003), Equivalence in a Cross-National Context:
Methodological and Empirical Issues in Marketing Research.
Tilburg, the Netherlands: Tilburg University Press.
Voss, Kevin A., Donald E. Stem, Lester W. Johnson, and Constantino Arce (1996), An Exploration of the Comparability of
Semantic Adjectives in Three Languages: A Magnitude Estimation Approach, International Marketing Review, 13 (5), 4458.
Werner, Oscar and Donald Campbell (1970), Translating, Working
Through Interpreters and the Problems of Decentring, in A
Handbook of Method in Cultural Anthropology, R. Naroll and R.
Cohen, eds. New York: Columbia University Press, 398420.

Collaborative and Iterative Translation

43