Cross-Cultural Comparison of Focus Groups as a Research Method

Introduction The validity of inferences drawn from focus groups rests on the verbal interaction between the focus group moderator and participants. When the focus group method is applied to research studies conducted in languages other than English, researchers need to make cultural and linguistic adaptations appropriate for the target population to maximize the effectiveness of the focus groups. However, there is a scarcity of research literature examining how focus groups perform in non-English languages, especially in Asian languages. Prior studies on the use of focus groups in cross-cultural research have centered on cultural sensitivity issues, procedures, planning, practicalities, and logistics (e.g., Colucci, 2008). As Clarke (1999) pointed out, the assumption underpinning the focus group method is that individuals are valuable sources of information and can express their own feelings and behaviors. It follows that focus group participants must verbally express their thoughts and behaviors; thus, the use of language plays a central role in focus group discussions. Although the research includes extensive discussion of methodological issues related to applying focus groups in non-English speaking cultures (Halcomb, Gholizadeh, DiGiacomo, Phillips, & Davidson, 2007), it lacks a systematic investigation to compare how speakers of different languages express their views and opinions in focus groups using those languages. Because focus groups use guided group discussions to gain insight into a specific topic, it is critical to examine the extent to which focus group participants engage in the discussion through verbal expression.


Introduction
The validity of inferences drawn from focus groups rests on the verbal interaction between the focus group moderator and participants. When the focus group method is applied to research studies conducted in languages other than English, researchers need to make cultural and linguistic adaptations appropriate for the target population to maximize the effectiveness of the focus groups. However, there is a scarcity of research literature examining how focus groups perform in non-English languages, especially in Asian languages.
Prior studies on the use of focus groups in cross-cultural research have centered on cultural sensitivity issues, procedures, planning, practicalities, and logistics (e.g., Colucci, 2008). As Clarke (1999) pointed out, the assumption underpinning the focus group method is that individuals are valuable sources of information and can express their own feelings and behaviors. It follows that focus group participants must verbally express their thoughts and behaviors; thus, the use of language plays a central role in focus group discussions. Although the research includes extensive discussion of methodological issues related to applying focus groups in non-English speaking cultures (Halcomb, Gholizadeh, DiGiacomo, Phillips, & Davidson, 2007), it lacks a systematic investigation to compare how speakers of different languages express their views and opinions in focus groups using those languages. Because focus groups use guided group discussions to gain insight into a specific topic, it is critical to examine the extent to which focus group participants engage in the discussion through verbal expression.
Our purpose in this chapter is to conduct a systematic analysis of the linguistic behavior of speakers of five different languages to compare how active they are in focus group discussions. This chapter has two objectives: (1) to examine the conversational style of focus group participants across languages and (2) to outline the interaction patterns between focus group moderators and participants as well as among participants. The ultimate goal is to provide a general picture of differences and similarities across language and cultural groups in terms of participatory patterns.
More specifically, we used a coding scheme, based on sociolinguistic theory, to compare and contrast how speakers of five languages (English, Chinese, Korean, Spanish, and Vietnamese) participated in focus groups. Four to six focus groups were conducted in each of the five languages to evaluate the data collection materials planned for the 2020 US Census. Our findings will contribute to ongoing research on the effective use of focus groups as a method of studying the public opinions of culturally and linguistically diverse populations in the United States.

Use of Focus Groups in Social Science Research and Cross-Cultural Concerns
The purpose of a focus group is to generate group discussion to gather qualitative information about the group's beliefs, attitudes, and behaviors relating to an issue, product, or service. Use of focus groups increased among social scientists in the 1980s, and several textbooks on the subject appeared in the 1990s (e.g., Krueger, 1998;Morgan, 1997). Focus groups are now commonly used by market researchers, academics, nonprofit organizations, government agencies, and community organizations (Krueger & Casey, 2000). Survey designers have also used focus groups to help conceptualize, contextualize, and frame questions; identify appropriate terminology for respondents; and evaluate questions (e.g., Campanelli, 2008;Fuller, Edwards, Vorakitphokatorn, & Sermsri, 1993;Kaplowitz, Lupi, & Hoehn, 2004). Notably, nearly all of this research drew participants from the same language group.
Conducting research in languages that respondents prefer presents new challenges for focus groups. The basic assumption of the focus group method is that focus group participants are expected to verbalize their thoughts and express their opinions. In addition, they are encouraged to interact with one another and are not limited to answering the moderator's questions only.
Because of the high amount of language use in focus group discussions, differences in communication styles across language groups inevitably affect the level of interaction among participants.
Communication styles refer to the ways in which speakers of a language or members of a cultural group use language to interact with one another. Sociolinguistics scholars have long pointed out that systematic and observable differences in communication norms across different languages exist (e.g., Gumperz, 1982Gumperz, , 2001Tannen, 1984). For example, when comparing Peninsular Spanish and British English in debates broadcast on television, Ardila (2004) found that Spanish speakers tended to be uncomfortable with silence. Thus, Spanish speakers often interrupted to express agreement and took advantage of a pause to take the floor. Félix-Brasdefer's (2003) study compared directness in declining an invitation among three groups: Latin American speakers of Spanish (native), Americans speaking Spanish (nonnative), and Americans speaking English (native). Controlling for gender, education, age, and Spanish dialects, researchers noticed that the Americans speaking English were more direct than the Latin Americans speaking Spanish, while the Americans speaking Spanish exhibited an intermediate frequency of directness.
In addition, language and cultural scholars have concluded that Confucian-based collectivist cultures (e.g., China, Korea, Vietnam) place a high emphasis on face (such as honor, respect, and social status); therefore, it is important to use the appropriate terms of address and polite expressions or lexicons that enhance the other's face (Kádár & Mills, 2011). Observational studies show that the Korean language uses a highly developed system of address terms that have many honorifics. Using a wrong term of address in speaking is a social taboo (Kim, 2011). The Vietnamese language is similarhonorific and kinship terms as politeness markers are considered important features in conversation (Chew, 2011).
This concern for politeness often leads speakers of Asian languages to habitually use vague expressions or short answers in question-answer settings. Some evidence in survey research shows that there are differences between Western-and Asian-language speakers when responding to questions in a survey research interview. Chan's (2013) study showed that, compared with English speakers, a higher proportion of Chinese speakers provided indirect responses when asked research interview questions and when asked to participate in a survey. Pan's studies (2008Pan's studies ( , 2012Pan's studies ( , 2013 also demonstrated remarkable differences between English and Chinese speakers in cognitive interview settings. English speakers were expressive during the English-language interviews. Their answers were characterized by detailed comments on the issues being discussed and elaboration on their individual opinions and feedback. In contrast, Chinese speakers in the Chinese-language interviews tended to provide brief, vague, and ambiguous answers; sometimes the answers were unrelated to the topics being discussed. They also used a community-based argumentation style (a we-based versus an I-based style) and answered "yes" freely to every question.
Researchers have documented the challenges of using focus groups in non-Western languages and cultures. Halcomb et al. (2007) conducted an extensive literature review of focus groups in culturally diverse settings and provided some key considerations for researchers. One consideration is that the concept of power relationships differs in non-Western cultures. For example, they note that in some non-Western cultures "it is considered rude for younger persons to even suggest they have different opinions from those of an older person or one who is considered more 'senior' or 'important'" (Halcomb et al., 2007(Halcomb et al., p. 1003. Various aspects of culture may affect the degree and nature of interaction among focus group members as well. Huer and Saenz (2003) reported that cultural mistrust may negatively affect participants' willingness to disclose information. Extensive knowledge of the participants' cultures is considered essential for conducting focus groups successfully. For example, they note that some Vietnamese Americans experienced government persecution in Vietnam and, as such, may be unwilling to participate in research studies. Also, cultural mistrust may arise because of concerns about how members of the target population believe they are perceived by the larger society. Huer and Saenz (2003) noted that in focus groups on attitudes toward disabilities, many Vietnamese Americans qualified their answers because they wished to avoid negative stereotyping.
However, comparative studies examining focus groups across languages are limited. One notable study is that of Lee and Lee (2009), which reported comparisons between focus groups conducted in the Netherlands and South Korea. They drew on differences in communication styles between highcontext cultures (e.g., China, South Korea) and low-context cultures (e.g., the Netherlands, the United States). They hypothesized that members of lowcontext, individualist cultures have different attitudes toward discussion and conflict than members of high-context, collectivist cultures. For example, particularly in focus groups conducted in South Korea compared with the Netherlands, they found lower levels of interaction among focus group members.
We contribute to this line of scientific inquiry with a comparative study that examined cross-cultural differences in conversational styles and interaction patterns among participants drawn from different language groups.

Research Questions
Based on the literature on cross-cultural differences in communication styles between speakers of Western and Asian languages, we predicted that a similar pattern can be observed in their focus group participation. We took the approach of treating a focus group as a communicative event (Saville-Troike, 1989). A focus group, like any other communicative events, has its general purpose of communication, topics of discussion, participants, language variety, tone, and rules for interaction in the discussion. When participants enter into a communicative event, they draw on background knowledge acquired through past communicative experience to infer what was intended and to act based on their cultural norms of communication (Scollon & Scollon, 2001). This background knowledge includes their familiar way of talking and the communication style that is preferred in the situation. Communication style can be investigated through a systematic analysis of linguistic features that constitute what Tannen (1984) calls "conversational style." According to Tannen (1984), conversational style results from habitual use of linguistic devices motivated by the overall strategies of Rules of Politeness (Lakoff, 1973;Lakoff' , which serve basic human needs in interaction, that is, the need for rapport (high involvement) and need for distance (considerateness). The involvementconsiderateness dimension in conversation has shed light on research in cross-cultural communication (e.g., Gumperz, 1982;Tannen, 1980) because conversational style can be placed on a continuum of high involvement to high considerateness, which enables researchers to easily identify linguistic features that show a pattern of interaction (see Tannen, 1984, pp. 30-31, for a list of linguistic features of high-involvement style).
We borrowed Tannen's term high involvement in this study to refer to active participation in focus group discussion, such as volunteering answers, giving elaborate comments, and actively interacting with other focus group participants. We used the term low involvement in this study in contrast to high involvement. A low-involvement style is characterized by a lack of interaction among focus group participants or short or brief answers, silences, or pauses in discussion.
In our study, we used this tool of linguistic analysis of conversational style to answer three specific research questions that correspond to the study objectives: • Do speakers of the five languages show the same or different interaction patterns in focus group discussions?
• Are Western-language speakers (i.e., English, Spanish) more likely to use a high-involvement style in focus groups?
• Are Asian-language speakers (i.e., Chinese, Korean, Vietnamese) more likely to use a low-involvement style in focus groups?

Data and Methods
This section documents the data and methods we used, the decisions and assumptions we made in the data collection and analysis process, and ways we mitigated the limitations. This transparent documentation approach was guided by the quality standards for qualitative research described in Lavrakas (2013) and Roller and Lavrakas (2015).

Data for the Study
Data for this study were drawn from focus group discussions that were part of a research study conducted by the US Census Bureau. The objective of the overarching study was to develop and pretest census data collection materials in multiple languages to ensure that they were linguistically and culturally appropriate. The materials and moderator's guide were developed in English first and then translated using a team-based translation approach (Harkness, 2013;Pan, Sha, & Park, 2019) by language experts who worked directly and iteratively with protocol designers and subject matter experts. Experienced focus group moderators used a semistructured protocol to conduct 22 focus groups with 205 participants. Six focus groups were conducted in Spanish, and four each were conducted in Korean, Chinese, Vietnamese, and English. Each group had 8-10 participants. For each language, half of the focus groups discussed data collection materials associated with an Internet self-response instrument, and half discussed materials designed for use on the in-person census interviewer visits. The focus groups took place between June and September 2015, and the 2-hour discussions were audio-and video-recorded with the consent of the participants.

Moderators and Participants
The study used six moderators: one each for Korean, Chinese, and Vietnamese and three for Spanish and English. Each moderator was experienced in conducting focus groups in the assigned language and was familiar with the census materials because they had worked on developing the non-English versions. Although they may have differed in their moderating styles and group dynamics were not predictable, we minimized inconsistency across the groups by having moderators use the same moderator's guide to ask questions, follow the topic sequence, and manage the allotted discussion time. All moderators completed up to 3 hours of formal, study-specific training, except the Korean moderator and one of the Spanish moderators who were part of the team that designed the study protocol. To build rapport with the participants before the start of the focus group discussion, the moderator engaged them in an icebreaker exercise about their shared experiences living in the United States.
The participants received $75 as a token of appreciation for participating in the 2-hour discussion. The majority of the focus groups were conducted in dedicated facilities in California, Illinois, Maryland, and Florida, while three focus groups were conducted in a professional conference room in North Carolina (one English and two Spanish). To be eligible for the non-Englishlanguage focus groups, a participant had to speak Spanish, Korean, Chinese, or Vietnamese as their native language and also speak limited English. This homogeneity gave them the same frame of reference when thinking about translations and the US Census. The participants in the English-language focus groups had to speak English as their native or near-native language, and they discussed English-language materials.
The participants were recruited via word of mouth, or they saw advertisements about the study and contacted the recruiters. To achieve a wide range of opinions in the discussion, we recruited participants based on characteristics such as education, age, sex, and, if applicable, the year they came to live in the United States. These characteristics represent a crosssection of the Spanish and Asian language speakers in the United States and reflect the authors' years of experience conducting research with these populations. For example, Koreans tend to have higher education, so recruitment of people based on educational attainment focused on high school and college graduates for the Korean focus groups. For the Spanish and Chinese focus groups, we also recruited participants from different origins to enrich the discussion about translations. For example, while there is a degree of universality in Spanish and in Chinese, there are differences in word use among people of various Spanish and Chinese origins. We did not intend to use these demographic and respondent characteristics as units of analysis because recruitment for qualitative research does not render a high enough number of cases to enable analysis by specific characteristics. Table 8-1 summarizes the composition of the specific groups.

Transcription Process and Verification
The focus group discussions were recorded and transcribed in the languages in which the focus groups were conducted. To ensure a level of consistency across the transcriptions, the transcribers were trained to type all utterances from the video recordings. They also followed a set of 15 transcription rules designed by the authors to indicate the speech pattern, such as stress (grammar), intonation (falling, rising, and continuing), pause, and laughter and nonverbal behaviors. The transcription was read by the moderator or a lead researcher to verify its accuracy.

Coding Scheme
We developed a coding scheme using the basic principles in linguistic analysis of conversational style (Tannen, 1984). We considered the interactions that take place in a focus group discussion (e.g., moderator-to-participant and participant-to-participant interaction) and the setup of such interactions (e.g., question-answer format and group setting). More specifically, we coded five distinct linguistic features to identify focus group participants' interaction patterns: participants' responses to moderators' questions, interaction direction, overlapping speech, and types of answers to the questions. Altogether, the four features have eight codes: interactions were labeled as voluntary, involuntary, participant oriented, moderator oriented, or overlapping, and each utterance from an interaction was coded as brief, elaborated, or back channeling. Table 8-2 shows the coding scheme with a description of the codes, their definitions, and objectives. After we developed the draft coding scheme, we piloted it on a small sample of focus group transcripts. The results suggested that the coders needed more specific instructions and examples, so we provided individual coaching.

Spanish
The participants represented origins from Mexico (n = 12), Central and South America (n = 18), and the Caribbean (n = 9), and two of the six focus groups were conducted with Puerto Ricans who lived stateside (n = 19).

Korean
Participants were grouped by age to minimize the seniority effect in the Korean culture that would affect group dynamics: two "younger" groups 18-44 years old (n = 19) and two "older" groups 45 years or older (n = 19).

Chinese
The written materials were in simplified Chinese, and the moderator and participants used Mandarin. Participants represented the major dialects of Mandarin, Cantonese, and Shanghainese and Chinese-speaking regions including China (n = 22), Taiwan (n = 5), and Hong Kong (n = 8).

Vietnamese
No additional specific characteristics were recruited.
a Nineteen of the 26 participants were Puerto Ricans who had lived stateside since 2010 because of the Census Bureau's research needs.
Each linguistic feature signaled a certain characteristic on a high-vs. low-involvement dimension. A high-involvement style is characterized by the participant's voluntary participation in the discussion, elaborated answers to probing questions, and multidirectional interactions (moderator to participant, participant to moderator, and participant to participant). A low-involvement style is characterized by the participant's involuntary participation (being called on), brief responses to probing questions, and single-directional interactions (moderator to participant). By examining these linguistic features, we compared and contrasted the conversational styles and interaction patterns across groups.
The unit for coding is a speaking turn taken by a speaker. A speaking turn is defined as the speech that a speaker produces without interruption from other speakers. A speaking turn can be as short as one word (e.g., "yes," "okay") or as long as several lines. Table 8-3 gives an example for each code  from the data to further illustrate how the coding scheme works. Each speaking turn has at least three codes (up to four codes including "overlapping").

Coding Process and Verification
We selected one segment from each transcript for coding. For the transcripts from focus groups that reviewed self-response data collection materials, we selected the segment that discussed a multilingual trifold brochure that was printed in color. For the transcripts about the materials associated with in-person census interviewer visits, we selected the segment that discussed a video clip showing an interaction between the interviewer and the respondent. These segments were both at the beginning of the group discussions. We selected them rather than coding the entire transcript to better manage and monitor the accuracy of the coding across the five languages. We decided not to select later segments because there might be potential bias in the interactions about recurring translation issues (i.e., identical translation issues that appear more than once): the participants may not state their opinions again (or may shorten them), and the moderator was not trained to probe on recurring translation issues because it would be repetitive. For each language in the transcripts, two coders completed the coding. They were part of the language expert panels assembled for the study that developed the materials for the focus group discussions but did not moderate the focus groups. The coders received 4 hours of training, including 2 hours of group training about the research objective, the coding scheme, and the procedure for documenting appropriate codes at each utterance, followed by instructions on using an Excel program for tallying. They also completed 2 hours of coding exercises at home and received feedback from the lead researchers.
The coding steps were as follows: (1) Using the same focus group transcript, both coders coded the same sections and then compared the codes. One of the coders was responsible for indicating discrepancies, making notes, and compiling results. (2) Coders consulted with lead researchers on the intent of specific codes and clarification of coding rules if there were discrepancies. (3) Coders for each language subsequently met in one or more meetings to reconcile any discrepancies.
After the first two coding steps, the Spanish and Vietnamese language coding did not reach 90 percent agreement, while the English, Chinese, and Korean language coding did. For all languages, the coders met once or more to reconcile the discrepancies and reach 100 percent agreement. For Spanish and Vietnamese, the reconciliation meeting revealed that the discrepancies were primarily due to the process: (1) the Spanish language coders' inconsistent handling of the transcripts (e.g., unsure where overlapping comments started and ended) and (2) confusion about specific coding rules. The reconciliation meeting between coders of each language ultimately resolved the discrepancies, and the coders reached an agreement on all codes. Table 8-4 documents the agreement by language before and after the reconciliation meetings.
For this study, we did not use complex statistics to evaluate coder agreement because the coders facilitated intercoder agreement by reconciling coding discrepancies through discussions. The coders were also quite knowledgeable about the subject matter, which reduced the likelihood that their coding agreement occurred by chance rather than as a result of actual agreement between the coders.
In summary, the focus groups were conducted in five languages, and many variables could not be controlled (e.g., unpredictable group dynamics). We attempted to mitigate them by using a consistent approach: experienced moderators used the same protocols and had a common understanding of the research objectives through training or roles as the protocol designers. The transcribers and coders followed a set of standardized procedures and verifications. In addition, the participants were homogeneous in terms of shared native language and limited Englishlanguage proficiency. All groups reviewed the same content of materials and videos and, in general, did not differ greatly in demographic characteristics. Because the strength of the focus group method lies in its qualitative, explorative nature, the flexibility and focus on context in the group discussions make it difficult to render the data absolutely accurate or inaccurate like in structured quantitative data collection. By fully disclosing the group compositions and our consistent data collection and analysis process, we hope the reader is enabled to reach decision that the comparisons across groups in this study are valid.

Findings
To explore the conversational styles of focus group participants in the five language groups, we took two steps in our analysis. First, we examined the frequency of occurrences of each linguistic feature in each language group to get an overall interaction pattern. We then compared and contrasted the interaction patterns among the five language groups to identify similarities and differences in those patterns. Second, we conducted a qualitative analysis to explore salient points identified in the overall pattern and to provide context for the main departure from the communication norms found in the analysis. Our findings address interaction patterns (Research Question 1) and involvement styles (participatory patterns) of Western and Asian speakers (Research Questions 2 and 3).

Quantitative Analysis
To address the research questions, we examined the frequencies of utterances by each code. There were 1,824 utterances and 22 groups in the analyses. Because group dynamics were not identical and inaudible utterances in the transcripts were coded as missing, the number of utterances in each analysis was different. As shown in Figure 8-1, English focus group participants had the highest percentages of voluntary responses, with 99 percent of the responses being  When we analyzed participants' utterances by interaction direction (moderator vs. participant orientation), we found that Spanish-speaking participants showed the highest level of participant-oriented interaction (46 percent), followed by English (40 percent), Korean (38 percent), Vietnamese (24 percent), and Chinese (14 percent) (see Figure 8-1). According to the coding scheme (see Table 8-2), participant-oriented interaction is mutually exclusive to moderator-oriented interaction. This means that Chinese and Vietnamese focus group participants showed the highest level of moderatororiented interaction rather than responding to other focus group participants' comments. Again, the result shows a strong participatory pattern for the English, Korean, and Spanish groups. The Chinese and Vietnamese groups had a weaker participatory pattern.
Overlapping speech can also reveal the interaction patterns of the participants. As indicated by Figure 8-1, among the five language groups, the Korean focus group participants had the highest level of overlapping speech (25 percent), followed by Spanish (24 percent), English (17 percent), Chinese (13 percent), and Vietnamese (11 percent). These findings indicate that the Korean, Spanish, and English groups tended to be more involved by overlapping speech, which is also a sign of a stronger participatory pattern. Figure 8-2 illustrates that, in terms of types of answers, the Korean focus groups had the highest level of brief answers (47 percent), followed by Chinese (40 percent), Spanish (25 percent), English (22 percent), and Vietnamese (16 percent). Vietnamese focus groups had the highest level of elaborated answers (81 percent), followed by Spanish (73 percent), English (72 percent), Chinese (52 percent), and Korean (50 percent). Chinese and English focus groups back channeled on a similar level, at 8 percent and 6 percent, respectively. Korean, Vietnamese, and Spanish focus groups back channeled similarly between 2 percent and 3 percent. Figure 8-2 illustrates a strong participatory pattern for the English and Spanish focus groups because of their elaboration and back channeling. Although Chinese and English groups back channeled at a similar rate, Chinese groups did not provide elaborated answers and therefore had a weaker participatory pattern in this analysis, along with Korean groups.
Vietnamese groups exhibited a strong participatory pattern that was driven by being elaborate in expressing opinions but maintained a weak participatory pattern in prior analyses.
Next, we conducted further analyses to investigate how each language differs from one another in terms of their linguistic features. The results in Table 8-5 indicate that for every linguistic feature, the differences between each language are statistically significant at the .05 level. The use of statistical tests to interpret coded qualitative data collected in focus group discussions does not differ in spirit from how researchers quantify recorded human communications in content analysis (Krippendorff, 2013, pp. 194-199).
As shown in Table 8-5, English focus groups were significantly different at the .05 level from the other four language groups in terms of voluntary responses made. Looking at the percentages, the Chinese and Vietnamese focus groups provided voluntary responses less frequently than any other languages, and these two groups were also significantly different from the Korean and Spanish focus groups in the pairwise comparisons of voluntary responses.
The Chinese and Vietnamese focus group participants were also less frequently engaged in participant-oriented interaction, and they were significantly different from the Spanish, English, and Korean focus group participants who demonstrated participant-oriented interaction more frequently. In terms of overlapping speech, Chinese and Vietnamese focus group participants overlapped less frequently than Korean and Spanish focus group participants, and the differences were significant. However, the Spanish focus groups were not significantly different from the English focus groups on overlapping speech.
For the level of elaborated answers, the Vietnamese focus groups were significantly different from the Korean and Chinese focus groups. Vietnamese focus groups were not significantly different from the English and the Spanish focus groups, which provided elaborated answers more frequently than the Korean and Chinese groups. The Korean and Chinese groups were not significantly different from each other, but they were different from the other language focus groups and provided elaborated answers less frequently. Lastly, the English and Chinese focus groups were not significantly different in back channeling. These two groups back channeled more frequently, and they were significantly different from the Spanish focus groups that back channeled less frequently. Although the Korean and Vietnamese focus groups were significantly different from the Chinese focus groups, they were not significantly different from the English focus groups.

Qualitative Analysis
The qualitative analysis supports the quantitative findings that the English and Spanish focus groups had similar interaction patterns of a highinvolvement style. They were characterized by more voluntary responses to the moderator's questions, more interaction among participants, and more elaborated answers in the discussion compared with the Chinese and Vietnamese focus groups (except Vietnamese for elaborated answers). Table 8-6 from an English focus group exemplifies the participants' active participation in the discussion. In this segment of the discussion, the moderator requested the group's reaction to a multilingual brochure that asked respondents to participate in a census test. The moderator showed the group the multilingual brochure and asked for their impressions regarding the placement of multiple languages in the brochure. As shown in Table 8-6, the first noticeable feature was the high level of overlapping speech. For example, when the moderator commented on the layout of the brochure (lines 502-504), Participant P9 started talking before the moderator finished his comment and gave an elaborated answer (lines 506-508). Second, multiple participants took part in the discussion (P3, P6, and P9), and there was  negative … but one thing that I did want to mention was that, French? It's very 530 common language, I can understand that they can't put all 5,000 languages on the 531 card, same as they can't put all 5,000 languages on instructions for a product that 532 you purchase, but I've always seen French. And I speak French and you know I 533 think it's just as common as Spanish. (V, E, P) overlapping speech between P9 and P3 and the moderator and P9. Participants' responses were all voluntary and were elaborated in most instances. Finally, the discussion was multidirectional, with interactions between the moderator and participants and among participants. The interaction pattern of the Spanish focus group is similar to that of the English group. (Due to space limitations, we only look at an English example here because the Spanish example had similar results.) Compared with the English and Spanish focus groups, the Chinese and Vietnamese groups showed a low-involvement style or weak participatory patterns, which are characterized by more involuntary responses to the moderator's questions, more interactions between the moderator and participants (than between participants), and briefer answers. (Chinese and Vietnamese groups shared similar interaction patterns except for one feature-elaborated answers.) Table 8-7 shows that in the Chinese focus group discussion, the moderator had to call on participants to provide feedback, and the participants' responses were brief. The number of involuntary responses was in sharp contrast to the English-language interactions. For example, in lines 552-553, the moderator tried to ask for a volunteer to respond. When no one volunteered, he urged the group to hurry up and speak and then called on P9. In this short segment, the moderator called on four participants (P9 in line 553, P4 in line 557, P8 in line 565, and P7 in line 571). In addition, the interaction was between the moderator and participants only. There was no interaction among the participants; they simply answered the moderator's questions and did not make comments on one another's responses. Most of their responses were brief. All these features suggest a weak participatory tendency for the Chinese-language focus groups.
The Vietnamese-language focus group differed from the Chinese group in one feature: types of answers. While the Chinese focus groups had the second lowest percentages of elaborated answers shown in Figure 8-2, the Vietnamese interactions identified the highest frequency of elaborated answers. As shown in Table 8-8, the moderator asked a simple yes or no question, but P11 volunteered an elaborate answer, stating why the material under review was easy to understand. This participant was a younger, more recent immigrant. Other instances of elaborated answers in the Vietnamese focus groups were also made by younger, more recent immigrants.
Among the three Asian-language focus groups, the Korean-language interactions showed the highest involvement and were the liveliest. The Korean focus group members voluntarily participated in the discussion at a much  higher rate than the Chinese and Vietnamese focus group participants in terms of their frequency of voluntary responses, participant-oriented interactions, and overlapping speech. The Korean groups also resembled English and Spanish groups when making voluntary and participant-oriented responses and had the highest overlapping rate of any language. Table 8-9 from a Korean focus group discussion demonstrates the group's higher involvement. In this excerpt, P10 voluntarily initiated a comment about the design of the multilingual brochure without being called on (lines 704-705). While P10 was still speaking, P6 indicated her agreement and added more points (line 707), also without the moderator's prompting. Another participant (P11) pointed out an observation that the other participants did not mention (line 711), and the moderator gave her feedback in line 713. The interactions continued when P11 clarified the point in line 715. Then, P3 (line 719) asked P11 a question, and finally, P9 wrapped up the whole conversation with a concluding remark in lines 723-724. In this short excerpt, we can see that the conversation was lively and included six people (five participants and the moderator). The finding that the Korean focus group discussions showed higher involvement and livelier participation is unexpected because it does not conform to the typical communication pattern of Asian languages as discussed in research literature (Lee & Lee, 2009). We attribute this unexpected finding to two factors. First, all the Korean focus groups were moderated by one of the lead researchers who designed the study protocol. She readily clarified points of confusion and flexibly guided the flow of the conversation. In comparison, the moderators of the other Asian-language focus groups followed the moderator's guide more closely. They were similarly  experienced in focus group moderation and familiar with the census materials, but they did not have the advanced knowledge of the study objectives like the Korean focus group moderator. Second, the Korean focus groups were divided into two groups-a younger group (aged 18-44) and an older group (aged 45 or older). This methodological consideration reflected the Korean cultural orientation on emphasizing varying expressions of politeness according to social hierarchy and respect for elders (Kim, 2011). The combination of subject matter expertise and culturally appropriate group composition likely fostered rapport among the participants and between the moderator and the participants. As a result, the discussion was livelier than it might have been otherwise.

Discussion
We conducted both quantitative and qualitative analyses to illustrate focus group participants' linguistic behaviors. The systematic analyses indicate that Western languages (English and Spanish) demonstrated similar interaction patterns. Asian languages (Korean, Chinese, and Vietnamese) shared patterns in many interactions, but the moderator's subject matter expertise and a culturally appropriate group composition could change that pattern (as was the case in the Korean group). In general, Western language speakers were more likely to use high-involvement styles and strong participatory patterns in focus groups than Asian-language speakers. Our study also demonstrates that each language has specific cultural dynamics and notable differences in focus group interactions. These interactions ranged from somewhat different to very different, and focus groups using the same language did not always exhibit very similar conversational styles. These findings reflect the dynamic nature of focus group data collection (which is also a strength that researchers rely on to interpret dynamic human interactions). Focus groups can still be an effective method for conducting research across cultural and linguistic groups when inherent sources of variability are mitigated by using a consistent data collection and analysis process and fully disclosing the details (see Data and Methods section).
In our experience, the efficacy of focus groups increases when the researcher develops strategies to address the factors that may affect group dynamics. For example, we recommend designing open-ended focus group probes (e.g., questions starting with "why," "when," and "what") to encourage more voluntary and elaborated answers and taking advantage of nonverbal cues (e.g., raising hands to show agreement) to facilitate group discussions. This way, there is sufficient information from a variety of participants to assess or inform the design of the data collection materials that are being discussed. In addition, an experienced and charismatic moderator with in-depth knowledge about the discussion topic and materials can encourage discussion while attending to cultural barriers and language nuances in conducting the focus group. Further research should be done to evaluate the efficacy of these strategies across language groups.
This study raised some important methodological considerations for conducting focus groups in non-English languages. The Korean interaction pattern in this study shows that with careful attention to group dynamics and methodological design (including moderator selection and training), researchers can obtain the desired participatory pattern in a non-Western language focus group discussion. Researchers also need to consider the factors of sex, age, group size, and possibly year of emigration to the United States to achieve the ideal group dynamics. For example, in the Vietnamese focus groups, younger, more recent immigrants tended to elaborate on comments when they spoke (while still having low involvement in terms of other linguistic features). The Vietnamese population in the United States has a different history than the Chinese and Korean populations. The earlier Vietnamese arrivals were refugees of the Vietnam War, but the more recent immigrants mainly consist of immigrants reuniting with relatives already residing in the United States (Rumbaut, 2007). Grouping the Vietnamese participants by the year they moved to the United States could have possibly created more homogeneity in the group and encouraged higher involvement in their interactions.

Conclusion
Focus group discussion is a communicative event governed by cultural norms of communication. The observable patterns of interaction across different language groups might affect the effectiveness of focus groups in gathering in-depth information from participants. However, the differences in interaction patterns can be minimized if researchers are aware of these differences and the interrelatedness of cultural norms of communication and interaction patterns. This study is an attempt to offer some insights into these differences and potential barriers in conducting focus groups in languages other than English. We propose two ideas for future research: (1) examine whether these differences affect the quality of data collected from focus groups and (2) explore ways of designing focus groups to address these differences across languages and cultures.
This study has several limitations. First, the focus group data were based on a purposive sample limited to the speakers of the five languages in several US geographic areas. It may be difficult to generalize the findings to the home cultures of the non-English-language groups. Second, the specific group characteristics may have contributed to the observed differences. In our design, we were not able to randomly allocate participants to language groups. We also did not use sophisticated statistical analyses to tease out issues related to speaking turns. Although our intention was to not force a qualitative study into a quantitative model, future research could explore the use of appropriate statistical modeling to interpret coded focus group data to study public opinion (e.g., similar to content analysis research). Doing so might enable deeper comparisons of the outcome of the discussions, such as whether groups that provide longer responses in fewer speaking turns (e.g., Vietnamese) offer insights about data collection materials the same way as language groups that have more speaking turns but keep their responses brief.