Does Language Matter? a Study of Bilinguals’ Responses to Questionnaires Framed By Language

ABSTRACT - The emergence of global markets has been marked by a parallel emergence of bilingualism. This poses a growing dilemma concerning the language in which marketing communications should be framed, as it seems likely that the use of a language could form a trigger that places a message recipient in a certain frame of mind. This is the issue addressed in this paper. Data collected from two matched groups of bilingual survey respondents in Singapore, distinguished by age, gender and social class, show that language frame does play a significant part in generating different responses to market research questions.

Citation:

Roger Marshall, Eliza Kang, Kang Wai Geat, and Kok Chew (2002) ,"Does Language Matter? a Study of Bilinguals’ Responses to Questionnaires Framed By Language", in AP - Asia Pacific Advances in Consumer Research Volume 5, eds. Ramizwick and Tu Ping, Valdosta, GA : Association for Consumer Research, Pages: 87-93.

Asia Pacific Advances in Consumer Research Volume 5, 2002      Pages 87-93

DOES LANGUAGE MATTER? A STUDY OF BILINGUALS’ RESPONSES TO QUESTIONNAIRES FRAMED BY LANGUAGE

Roger Marshall, Nanyang Business School, Singapore

Eliza Kang, Nanyang Business School, Singapore

Kang Wai Geat, Nanyang Business School, Singapore

Kok Chew, Nanyang Business School, Singapore

ABSTRACT -

The emergence of global markets has been marked by a parallel emergence of bilingualism. This poses a growing dilemma concerning the language in which marketing communications should be framed, as it seems likely that the use of a language could form a trigger that places a message recipient in a certain frame of mind. This is the issue addressed in this paper. Data collected from two matched groups of bilingual survey respondents in Singapore, distinguished by age, gender and social class, show that language frame does play a significant part in generating different responses to market research questions.

THE EMERGENCE OF A DILEMMA

The question of how well bilinguals alternate between their languages has been addressed in the literature, as has the difficulties of translation (Grainger, 1993). With regard to the latter issue, it is clear that however good a literal translation may be, it is also important toknow how an individual perceives the meaning of the words they use, as words can acquire different connotative meanings within two cultures and/or languages. Thus, a perfectly well-translated instrument that uses words with different affective meanings in two cultures could provide data that are the results of these different meanings rather than the factual differences sought by the researcher.

At an even more fundamental level, as early as 1964 Ervin found evidence that English-Japanese bilinguals use different concepts in each language when reacting to the same stimulus in a free association task. Cross-cultural psychologists have also suggested that survey results may reflect the way language or cultural conventions affect answers rather than the actual content of the questionnaires utilized (Marin, Triandis, Betancourt & Yashima, 1983). Triandis (1972) found that calling forth the second language of bilinguals can elicit the subjective culture and behaviors appropriate to that language group. Yang and Bond (1980) later proposed that ethnic affirmation could explain the odd fact that they found Chinese bilinguals responded in a more "Chinese" direction when answering a questionnaire in English. Their explanation is based on the assumption that an individual’s ethnicity itself becomes a key element of the social environment. In 1982, Yang and Bond further argued that respondents might have cross-cultural accommodation factors, where the giving of responses is dependent on what is appropriate to that culture. They concluded that ethnic affirmation was observed with the more ego-involving (important) items while cross-cultural accommodation with the less involving. In his later research, Triandis discussed these two possible explanations for different responses of bilinguals, and determined that social desirability explains the largest number of differences in response found (Triandis, 1980).

In spite of the obvious interest displayed by market researchers in the perils of translating questionnaires for international work, little research has focused on the deeper significance of bilingualism to consumer behavior. There are several aspects that could be considered here, including the response of bilinguals to advertisements and to research questionnaires framed in different languages. The research question addressed here is whether or not bilingual consumers’ responses differ if questions are posed in either one of their two languages.

Both products and language communicate symbolic meanings to consumers; members of a society transfer meaning from their culture to various objects through language (McCracken, 1986). Thus language attitudes will often be social attitudes towards an object as well (Appel & Muysken, 1987). The social values of a person help to shape that person’s beliefs, and a consumer’s beliefs go on to help shape his or her attitude towards products. What of consumers having two cultural backgrounds, each expressed by a different language and each espousing a separate set of values? By setting the questions used in this research within the frame of specific values held by Chinese and Westerners, within multi-ethnic Singapore, the framing effect in question is given every chance to show.

There is some work on the socio-economic aspects of bilingualism that has a bearing. Fishman’s (1972) research shows that style and language choices are not random, but follow socially determined rules. The author identifies five domains where use of language might follow social rulesBfamily, friendship, religion, employment and education. These factors are similar to the items we have included in the research questionnaire for their language sensitivity. What is suggested here, however, is different to Fishman’s work, in that this approach is socio-cultural rather than socio-economic. This means that not only might a different information processing route be taken (as suggested by, for instance Miller, Speece and Miller, 1996) but that a socio-cultural mental heuristic is triggered by the use of a specific language, rather than a language choice being triggered by a socio-economic situation. The focus of the research reportd here, then, is to investigate whether language choice forms a cue which triggers a particular set of values and attitudes in the individual’s mind. If this is the case, then there are profound implications for market researchers (and others) in bilingual communities.

The Republic of Singapore is ideally located to conduct research to address this question, as it has a large proportion of bilingual people. Actually, the country has four languages which are officially recognized; Tamil (spoken by the Indian community), Malay, Mandarin Chinese (hereafter "Chinese") and English. English is, by far, the most commonly used language for public purposes, but Chinese is most commonly used in private by the large majority of the (mainly Chinese) population. It should be noted too, that many Chinese Singaporeans have a mother-tongue that is some dialect of Chinese and some only speak English. Since English and Chinese are the languages used to study this problem here, the results will clearly not be directly applicable to bilinguals who speak in other languages. However, the principleBthat language does provide a trigger to invoke a cultural heuristicBshould generalize to any bilingual community. Thus the (null) research hypothesis:

"There is no difference in the answers of bilingual (Chinese and English) Chinese Singaporeans to questions posed in English or Chinese."

TABLE 1

DEMOGRAPHIC CHARACTERISTICS OF THE SAMPLE

RESEARCH METHOD

Sample

After screening to ensure bilingualism, random assignation to quotas yielded two matching sets of Singapore Chinese bilinguals. To determine the quota categories, consideration was given to the factors that might influence a predilection to think in one mode or another (Eastern or Western) other than language frameBit was decided that age group, social class and (possibly) gender are all potential influencing variables and should therefore be taken into account when selecting the sample, in order to minimize sample bias.

It is speculatively hypothesized that age differences will probably be significant, because values within a culture do tend to change over time. In particular, as a generation cohort passes, values tend to change. The respondents were thus sorted into three age groups, approximating three such generations, of above 40, between 25-40 and below 25 years of age.

Similarly, social classes are predicated upon differences in value systems, and thus could offer a possible confounding factor. However, social class distinctions are always problematic in that the basis for social stratification tends to vary across cultures and over time. In Singapore this is particularly true, as the national culture of Singapore is young, so class lines are yet to be heavily drawn. Thus respondents were selected simply by "Blue" versus "White" collar workers instead of the more standard Upper, Middle and Lower class divisions.

The authors could think of no compelling reason why gender should have any influence whatsoever in this circumstance. The gender variable was, therefore, used simply to maintain a balance in the sample.

Research instrument

Six topics, spanning values of freedom, individualism, filial piety, sex, religion and environmentalism were selected deliberately as they were, subjectively, considered to be sensitive to differences in eastern and western cultural values. Filial piety values are usually considered very important in eastern cultures while issues on freedom and individualism and environmental concerns are usually more important in the West. Similarly, there are stronger social concerns with respect to sexual issues (and discussion of them) in eastern cultures than in western.

The assumption is made that difference in religions in the East and est may be a cause of the moulding of different value sets. Insofar as Chinese language is associated with the Taoist and Buddhist religions and the English language with Christianity, then these values may prevail when keyed by the use of one language or the other. Similarly, we presume that for the same question on sex, bilinguals answering the questions in Chinese will be more conservative compared to those answering it in English. Thus, not only might language trigger different value sets, but might also trigger different modes of expressing those values. It matters not if such a bias is described as a response, a social acceptability or some other bias; all are considered here part of a cultural heuristic triggered in a respondent by being asked to think in a particular language.

A list of possible questions, based upon these values, was then developed through brainstorming. These were tested by trying them on a group of bilinguals to act as cultural anchors for our study. Eighteen items were selected from the list, based on an analysis of their clarity of meaning, commonly understood sentence structure, grammar and terminology. After a few final adjustments were made, in line with the pre-test feedback, the final version was translated into Mandarin, pre-tested again, then back-translated to test for technical translation difficulties. No anomalies were identified. Two versions of the questionnaire were created in each language, with the order of questions varied to avoid order bias. The items used are listed in Table 2.

Procedure

The large sample size of 2,591 was obtained with the assistance of undergraduate consumer behavior students of a local university in Singapore. Working in teams of either three or four, they were given specific instructions on the type of respondents to look for. Each team was asked to collect 26 responses using the English-language version of the questionnaire from one group, and another 26 responses using the Chinese version from a different group. The teams were then asked, for course credit, to analyze their own data, whilst the good data sets were assembled into a single data base. Scrutiny of the individual data sets and their analysis by faculty provided some confidence that the data was soundBwhere any doubt was felt about the set it was discarded. Each undergraduate team collected their data based upon one factor (age group, gender or social class), keeping the other factors approximately even. All respondents were of Chinese ethnic origin; care was taken to ensure their bilingual ability and the English or Chinese version of the questionnaire was given to them based on random assignation rather than in response to their request. Great care was taken to ensure that these two large groups (Chinese language 1307, English language 1284) are equivalent, so that any differences in their responses can then be reasonably attributed to the language medium of the questionnaire.

TABLE 2

LIST OF QUESTIONNAIRE ITEMS

The weakness of this between-groups method is that the matching of the two groups is essential. To further add to the measures described above, a large sample was considered preferable so that the effect of outliers would be absorbed. The alternative method, of using a within-group, repeated-measures design, was considered, but was rejected. It was decided that, if the repeated-measures design was implemented, the learning effects of such a process would pose a problem.

DATA ANALYSIS

Triandis (1980) highlights the relevance of emic (culture-specific) factors in determining an accurate representation of each culture, rather than an etic approach which makes assumptions of similarity between the cultures under investigation. This suggests the necessity for the separation of English and Chinese answers when analyzing the data. However, the subjects of our study are all from the same culture and two groups have ben selected for analysis based on their comparability rather than their differences. Thus the decision to adopt an emic approach as opposed to an etic is complicated. On the one hand, it could be considered that the data from the two groups should be treated as if they come from different cultural groups. In this instance, separate, within-group, emic analysis should be conducted before between-group comparisons are made. On the other hand, it can be argued that the whole point of this research is that the analysis is between members of one cultural group, not two; therefore it follows that an etic approach is more suitable. It was noted that some prior research (Yang & Bond, 1980) apparently avoided this dilemma by analyzing between-group differences in terms of comparing differences in response to each question rather than by attempting to group questions through factor or cluster techniques. In fact, taking a question-by-question approach is adopting a de facto etic stance; as an assumption is made that every question has equal relevance and identical meaning to both groups. Triandis (1980) points out that the emic and etic approaches are actually complementary and the use of both can be instructive. This advice has been heeded here and an etic approach discussed first, followed by an attempt at emic analysis.

TABLE 3

REVISED VARIMAX-ROTATED FACTOR LOADINGS

Etic analysis

An omnibus ANOVA, using the mean of all 18 variables as the dependent variable with language, age and gender as independent variables, shows a significant effect for language (F=19.951, p<.001). Further testing, by the Kaiser-Meyer-Olkin measurement of sampling adequacy (65%) and the Bartlett test of Sphericity (2754.7710, p<.001), confirm that the data set is appropriate for further inspection by factor analysis, in order to eliminate redundancy and reduce the variables to a more manageable number. Principal component analysis was utilized, followed by a Varimax rotation. Five factors emerged. Questions 1, 5, 10 and 16 failed to load uniquely onto any of the five factors and so were removed from further analysis. A further analysis was carried out and, again, five factors emerged, that together explain 49.7% of the variation in the data set. The revised factor loadings are reported in Table 3.

Naming the Factors

The first factor, consisting of Questions 8, 17 and 18, basically involves the issue of concern towards the environment; hence, the factor is named "Environmental issues" (see Table 3). It is slightly puzzling why Question 8 (which deals with parents consulting their children about issues concerning them) should load so heavily on this factor, when questions 17 and 18 deal so directly with concern for the environment. As always with factor analysis, there is a chance that the items correlate by chance, or in relation to some other item that may not even be included in the analysis. On reflection, it seems at least possible that parents with an enlightened, modern attitude toward environmental concern might also have a modern, non-traditional attitude towards consulting their children. The second factor, which involves Questions 2, 4, 7 and 9, revolves around the issue of filial piety, so is named accordingly. The third factor, "Religion", is formed by Questions 13, 14 and 15. This is a logical grouping as all three questions pertain to religious issues. The situation is similarly straightforward for the fourth factor. The two questions making up the factor, Questions 11 and 12, are concerned with the issue of the degree of openness about sex, so was named "Conservatism". Finally, factor five, which is made up of Question 3 and 6 is named "Freedom" because the questions aim to test the respondents on their autonomy and resistance to external influences.

A further ANOVA was run for each of the five factors, to gauge the significance of the effect of language on them. Five models were run, with the five factors taking turns to be the dependent variable while language, gender, age and social class were the independent variables. Two of the models, for religion and individualism, show ahighly significant main effect for language. This implies that there is a difference in the degree of sensitivity between respondents of the English and Chinese questionnaire on these two issues. This significance level, however, must not be taken at face value, as the sample is large and, therefore, it could be that the significance is merely a statistical artifact. To check for a problem here, Hay’s version of Omega squared (w²) is used. Cohen (1977) gives a guideline for categorizing small (w²⁼0.01), medium (w²=0.06) and large (w²=0.15) effect sizes (Cohen 1977). Thus the significance levels quoted in Table 4 do, in fact, represent real, but not large, differences. (w²_(Religion)=0.011, w²_{(Individualism)}= 0.008).

TABLE 4

SUMMARY OF ANALYSIS OF VARIANCES

TABLE 5

FACTOR STRUCTURE FOR CHINESE AND ENGLISH DATA SETS

A further consideration of the items that made up the factor "Concern" shows that it is actually made up of a variable on filial piety and two variables concerning environmental issues. It seems logical to remove the odd variable. This new variable, consisting of items 17 and 18, is named environmentalism. An ANOVA test is performed and a significant effect for language is obtained. (See Table 4). The effect size for this model is found to be w²_{(Environmentalism)}=0.001.

Emic analysis

In order to conduct an emic analysis it is necessary to perform a separate factor analysis for each of the data sets generated by the English and Chinese questionnaires to see if a comparable set of factors emerges. Oddly, the research hypothesis will be supported if the attempt fails. This is because it is the contention of the authors that each set of respondents will be prompted to think in a different way; the cognitive structure of the heuristic evoked by the culturally-laden question items will make factor structures quite different within the two groups.

This is, in fact, what transpires. Factor analysis, again using Principal Component analysis and a Varimax rotation, shows seven factors emerging for the Chinese data set, that explain 53% of the variance in the data set (eigenvalues range from 2.26 for Factor 1 to 1.02 for Factor 7). All items load satisfactorily onto a single factor, except for Question 6 which stands as a factor in its own right. Factor structure is reported in Table 5.

In the initial analysis of the English data two questions, (4 and 16) fail to load uniquely and were consequently removed from the set. Further analysis yields six factors that explain 50% of the variance (eigenvalues range from 2.09 for Factor 1 to 1.01 for Factor 6). Factor structure is also shown in Table 5.

Comparison of the two analyses does reveal some fundamental differences that do make comparison unrealistic. Some discrepancies can be reconciled up to a point. For instance, it seems as if religion and sex are separate issues for those questioned in English but, when prompted in Chinese, the matched group responds by classifying a woman’s virginity as a religious issue rather than a sexual one. Other differences are far more obscure and complex but certainly indicate that each factor illustrates a different cognitive structure attached to a specific, culturally-laden heuristic (factor).

DISCUSSION

This preliminary research has demonstrated that there is, indeed, an effect upon the responses of a bilingual audience caused by the frame provided by the use of one or other of the languages they speak. In a world where Chinese, English and Spanish are rapidly becoming the major languages spoken, bilingualism appears to be emerging more commonly. Thus is it typical that a European will speak a local language (German or French, for instance) as well as the English of the European Community. So, too, is it more and more common that Chinese nationals speak English, the dominant language of international commerce. It should also be noted that panish is now condoned as a language of instruction in some parts of the USA, whilst many Spanish-speaking people in South America also speak (American) English. The influence of colonialism is still felt, too, in the extent to which English is spoken in Great Britain’s former empire, and the historically appropriate language as well as the local language in the former territories of other ex-colonial powers such as France and Germany. Singapore is uniquely positioned to explore this issue, as the major national language is English yet the overwhelming majority of the population are of Chinese descent and still speak the language.

Business Implications

International business practice commonly demands communication to market segments described by factors other than national borders. This research shows that the picture may be even more complex than this. This is because wherever there is a significant sub-culture, or wherever bilingualism is common, then care will have to be exercised to establish if the communication topic is sensitive to language cues. This study has taken a first step in finding whether language can, indeed, affect bilinguals of the same market segment differently, and the results show positively that there is a difference in the answers of bilingual Singaporean Chinese to questions posed in English and Chinese.

In Singapore it seems as if (inter alia) religion, individualism and environmentalism are sensitive to language frame. Unfortunately, this list may not even be exhaustive within Singapore and certainly will be different in other bilingual situations. Nevertheless, the fact remains that there is a strong likelihood of some subjects being sensitive to language in any situation; therefore it behooves businesses and business researchers to check before making their communications.

The situation chosen here to test the research idea was that of responses to marketing research questions. There seems no reason, however, why the findings should not be generalized, in principle at least, to other forms of communication. Advertising is one such other form that springs to mind. Suitability of an advertisement, or the interpretation of the content, may well be language-dependent within a bilingual situation in just the same way as the research questions were here. For example, an organization like the Society for AIDS Prevention could try to prick the guilt of husbands endangering the family unit in Chinese, rather than in English, since the Chinese language medium can trigger the heightened family values sensitivities of these people. Again, advertising an environmentally friendly product in Singapore would probably be more effective in English than Chinese.

Implications For Theory

This work is only preliminary, and much remains to be done with regard to the details of exactly what areas are sensitive to language. Nevertheless, the principle has face validity in that what we know of memory structures suggests that clusters of associations (heuristics) are often triggered by some cue.

Other authors have also suggested that further research be carried out to explore the ethnicity of the experimenter (Triandis, 1983) and examine in detail the connotations of the words used in each questionnaire, including factors such as order of presentation, the context in which they are presented, and so on (Schuman & Presser, 1981).

The relative importance of various biases causing the differential response of bilinguals to questions in either of their languages is certainly a worthy research topic. On the other hand, all such biases are caused by thinking in a cultural-specific way about certain value-laden topics. It is not a long stretch of the imagination to consider that the complex, very large bundle of cultural information and values within an individual is bonded together into what could be considered a large, powerful cognitive heuristic. When a person is introduced to a new culture, as an immigrant perhaps, or lives as a member of an ethnic subculture in some community, then it may well be that a single individual could develop two such cultura heuristics. Thus a Chinese youth in modern Singapore will speak English at school and university, read English-language text books and watch English-language films and television shows; these all encapsulate a set of Western values. On the other hand, the same youth may have parents who prefer to speak Chinese in the home and may hold to an Asian set of values far removed from the Western values their children are exposed to outside the home. Thus the youth, faced with an English-language communication, will immediately be cued to a particular set of values that could be quite different, in some aspects at least, from those cued by a Chinese-language communication that contains exactly the same content. Although far from proven here, the logic of this argument is, we believe, compelling, and the topic well worth pursuing.

REFERENCES

Appel, Rene and Pieter Muysken (1987). Language Contact and Bilingualism. Edward Arnold (Publishers) Ltd., London.

Cheng, Chin-Chuan (1982). "Chinese Varieties of English". In Braj B. Kachru (Ed.), The Other Tongue. University of Illinois Press, London.

Cohen, Jacob (1977). Statistical Power Analysis for the Behavioral Sciences. New York: Academic Press.

Corson, David (1995). Using English Words. Kluwer Academic Publishers, Dordrecht.

Fishman, J. A. (1972). Language and Nationalism. Newbury House Publishers, Rowley, Mass:

Grainger, Jonathan (1993). "Visual Word Recognition in BilingualsBThe disruptive effects of switching language". In Robert Schreuder, Bert Weltens (Ed.), The Bilingual Lexicon. John Benjamin Publishing Co., Amsterdam/Philadelphia.

Kroll, Judith F. (1993). "Accessing Conceptual Representations for Words in a Second Language". In Robert Schreuder, Bert Weltens (Ed.), The Bilingual Lexicon. John Benjamin Publishing Co., Amsterdam/Philadelphia.

Lee, Kok-Cheong (1983). Language and Language Education. Singapore University Press, National University of Singapore.

Miller, Chip, Mark Speece and Birgit Miller (1996). "Chinese Information Processing and its Implications for Advertising Design". Proceedings of the International Management Development Association, Bermuda, July, pp 441-444.

Miller, Chip, Jim Reardon, Mark Speece and Birgit Miller (1966), "The Effects of Linguistic Styles on Cognitive Processes: Alphabetic vs Morphemic Processing", Proceedings of the Atlantic Marketing Association, v. XII, October, pp364-369.

Marin, G. Triandis, H.C. Betancourt, H. & Kashima, Y. (1983). "Ethnic affirmation versus social desirability: Explaining discrepancies in bilinguals’ responses to a questionnaire". Journal of Cross-Cultural Psychology. 14, 173-186

McCracken, Grant (1986). "Culture and Consumption; A theoretical account of the structure and movement of the cultural meaning of consumer goods" Journal of Consumer Research, 13, June, p.72.

Platt, John T. (1977). "A Model for Polyglossia and Multilingualism (with special reference to Singapore and Malaysia)". Language in Society 6: 361-78.

Schuman, H., & Presser, S. (1981). Questions and answers in attitude surveys: Experiments on question form, wording, and context. New York: Academic Press.

Singh, P.N., S.C. Hwang, and G.C. Thompson (1962) "A comparative study of selected attitudes, values, and personality characteristics of American, Chinese, and Indian students." Journal of Social Psychology 57:123-132

Triandis, H. C. (1972) The Analysis of Subjective Culture. Nw York: Wiley.

Yang, K. S., & Bond, M. H.(1980) "Ethnic affirmation by Chinese bilinguals". Journal of Cross-Cultural Psychology. 11, 411-425.

----------------------------------------