Backchannel Responses and Enjoyment of the Conversation: The More Does Not Necessarily Mean the Better

This study examined the types of backchannel response as well as its relationship with speaker presentation, listener recall, and participants’ perceived enjoyment of the conversation in an intercultural setting. Participants were 40 Anglo-Canadians and 40 Mainland Chinese, forming 40 same-gender dyads and performing two dialogues. All interactions were video-taped and micro-analyzed. Noteworthy findings include the following: 1) The Chinese participants in the role of listeners made significantly more backchannel responses than their Canadian counterparts in performing Task 2. 2) “Nod” and “okay” had the highest frequencies in both cultural groups. However, the Canadians used “repeat” more frequently than Chinese and the Chinese used “uhm” and “yeah” more than the Canadians. Participants in both groups “switched codes” when making backchannel responses, providing support for communication accommodation theory. 3) A significant negative correlation was found between the frequency of backchannel responses and participants’ self-reported level of enjoyment of the conversation, raising the critical issue of how to balance the appropriate amount of backchannel response in intercultural communication.


Introduction
Backchannel responses are common occurrences in conversations regardless of culture and language (Clark & Wasow, 1998;Goodwin, 1986;Heinz, 2003;Schegloff, 1982). Cultural and linguistic groups share some backchannel response codes and differ in others. Bilinguals switch from one set of codes to another just as they switch from one language to another. For example, when two Chinese converse in Mandarin Chinese, they use "shi ma", a rhetoric question to show that "I am paying attention to what you are saying". However, when a bilingual Chinese talks in English with a Canadian, "shi ma" is never used. Instead, "uhms" and "okays" are used, an indication that the Chinese converges to the backchannel responses of the English-speaking Canadian (Li, 2006).
Although patterns of backchannel responses of bilingual Mandarin-speaking Chinese have been studied, their functions have rarely been examined. One study reported a negative correlation between the frequency of backchannel responses and content communication in conversations between Chinese and Canadians (Li, 2006). The more backchannel responses were made, the less content information was transferred. On the other hand, when two Chinese conversed in Mandarin, more backchannel responses meant more content communication. As well, when two Canadians conversed in English, more backchannel responses also meant more content communication.
In the 14 categories of backchannel responses, Li observed that in both the Canadian/Canadian and Chinese/Chinese conversations, the majority of backchannel responses were in the categories of 'nod', 'nod with ok', and 'nod with uhm'. Canadians used 'nod with yeah' and 'okay' more frequently than Chinese whereas Chinese used 'uhm', 'right' and 'repeat' more often than Canadians. The Chinese also had a special category 'is that so' (shi ma) which the Canadians did not have. Li (2006) also noted that a two-way code switching in the two intercultural groups occurred, providing evident support for communication accommodation theory which states that interlocutors have a tendency to converge or diverge their linguistic codes in their conversations (Giles, Bourhis & Taylor, 1973;Giles, Mulac, Bradac & Johnson, 1987;Giles & Smith, 1979). When conversing with a Chinese, a Canadian used more "uhm" and "right" than talking with another Canadian. Similarly, when conversing with a Canadian the Chinese used fewer "right" and more "uhm" than talking with another Chinese. These findings are in agreement with a report by Tao and Thompson (1991) that native Chinese who were fluent in English had a tendency to switch code, using English backchannel responses.
In terms of the functions of backchannel responses, Li's study seems to be the only one which found strong positive correlations between the frequency of backchannel responses and listener recall scores in Chinese/Chinese and Canadian/Canadian conversations, an indication that backchannel responses played a positive role in transmitting the information from the speaker to the listener. On the other hand, she found significant negative correlations between the frequency of backchannel responses and listener recall scores in Chinese speaker/Canadian listener and Canadian speaker/Canadian listener interactions, showing that backchannel responses may have inhibited content communication. Could frequent backchannel responses be misleading signals in intercultural communication? In the present study, we continued to examine this issue with a larger sample size (40 intercultural dyads) than Li's 2006 study (20 intercultural dyads). In Li's 2006 study, no training was provided to the participants. In the present study, we provided a 15-minute training session to half of the 40 dyads making sure they felt comfortable to ask questions if understanding was in question. The research questions (RQ) were: 1. Were there any significant differences between the means of backchannel responses in the experimental and the control condition, in the doctor and patient roles, and in the Canadian and Chinese participants?
2. What were the characteristics of backchannel responses by speaker and listener roles?
3. What were the characteristics of backchannel responses in the two cultural groups?

Sample
Ninety-four university students participated in the present study. The participants formed 47 dyads, seven of which were eliminated from data analyses due to incomplete data or lack of fit to the criteria. According to the sampling criteria, all Caucasian participants must be born in Canada and speak English as their first language. All Chinese participants must be born in China and speak Mandarin Chinese as their first language. Chinese participants who had been in Canada for more than 8 years were not eligible. Both Canadian and Chinese participants must be under 35 years of age and they must not be a psychology major.
Among the remaining 80 participants, 40 were mainland Chinese (20 males and 20 females) and 40 were Anglo-Canadians (20 males and 20 females). The mean age for the Chinese group was 24.95 and that for the Canadian group was 23.73 years. These means were not significantly different from each other. Students were recruited in classrooms and university cafeterias, and through postings on the university bulletin boards. To ensure that the Chinese participants had sufficient English language ability to participate in the conversations, they were required to have achieved a university English proficiency level for reading and listening comprehension as demonstrated by their scores in the Test of English as a Foreign Language (TOEFL). All Chinese participants had TOEFL scores of 570 or above. At the time of the experiment, the Chinese participants had resided in Canada for an average of 4.01 years. Chinese students in the experimental and control groups did not differ in the number of years in Canada.
In their first encounter with the researcher, participants were informed of the nature of the study (i.e., a simulated medical interview) and that their conversations would be videotaped. Upon their arrival at the laboratory, participants were again informed that their conversations would be videotaped and that they could view their own tape afterwards if they wished to do so. Prior to giving instructions about the study, written consent was obtained from each participant regarding the way(s) in which the videotapes might be used.

Experimental Design and Procedures
A between-subjects design was used. The experimental condition had two intercultural combinations: Canadian physician/Chinese patient and Chinese physician /Canadian patient. The control condition had the same intercultural combinations: Canadian physician/Chinese patient and Chinese physician /Canadian patient. The decision for not including intra-cultural conditions was based on findings from previous studies that intra-cultural dyads did not have as many problems communicating as intercultural dyads since both parties used their native languages and interacted with someone from their own cultural backgrounds (Li, 1999a). Therefore, the focus of the present study was intercultural dyadic discourse.
Participants were paired with a partner of the same gender; that is, men were paired with men, and women were paired with women. Allocation of the dyad to the experimental or control group was randomly determined at the time of the pairing. The role of the participants was also randomly assigned upon their arrival at the laboratory.
All dyads engaged in the same communication task, which involved simulating a physician-patient interview. The session was divided into two parts: 1) the patient presenting the case history to the physician; and 2) the physician giving the patient instructions on the use of codeine. The case history was borrowed from Li (1999a). The Instructions on Codeine was taken from Compendium of pharmaceuticals and specialties (1982).
Immediately after the dialogues, the participants filled out a questionnaire which consisted of 13 questions asking about their experience of the interaction. Responses to the questionnaire are reported in the Results section.
The experimental condition. Upon arrival at the lab, participants were placed in separate rooms so that they did not communicate among themselves regarding the content or the procedure of the study. After the roles of either a patient or a physician were assigned, the participant playing the role of the patient was given a case history to study. He or she was instructed to take as long as needed and remember as many details as possible. A multiple-choice test (as a manipulation check) was then given to the participant in the patient role to ensure that he/she had mastered the content.
Meanwhile, participants playing the role of the physician received a short training session on grounding strategies. They were given a written list showing five ways to request their patients to explain, or repeat, or reformulate a previously stated piece of information (Appendix A). After they read the list, the researcher rehearsed the questions with them until they had mastered the material. Once the learning was complete, the researcher queried whether or not they would feel confident and comfortable to ask these questions during the dialogue. If a participant was hesitant, the researcher again reviewed the materials and gave assurance that it was all right to ask their patients questions whenever necessary. The training process lasted 10-15 minutes.
On the same page, was also a list of information that the participants in the physician role should obtain from their patients during their interactions. The list of information was relevant to a general physician-patient interview (e.g., an exact description of the problem, whether or not the patient had previously encountered the problem), and information on the list was not specific to the content of the case history.
The dyads were then instructed to engage in the conversation in a "talking manner." To minimize memory error, the patient was allowed to refer to the case history sheet while engaging in the conversation, but was not permitted to read from it word for word. Afterwards, the participant with the role of physician took an open-ended test to measure how much information related to the case history was successfully communicated.
Thus, the first task was completed. Before participants started the second task (physician gives instructions for the use of codeine), the participant playing the role of the physician was given time to study Instructions for Codeine while the patient received training on grounding. The procedures were identical to Task 1 except that the patient now received training while the physician studied the instruction sheet. After their conversation, the patient took an open-ended test, which measured how much information about Instructions for Codeine was successfully communicated.
The control condition. Participants in the control condition followed the same procedures and performed the same tasks as participants in the experimental condition except that they received no training on grounding.
Immediately after the dialogues, the participants filled out a questionnaire which consisted of 13 questions asking about their experience of the interaction. Responses to the questionnaire are reported in the Results section.
The average time for participants to finish the two conversations was 620 seconds across conditions. The mean times were 662 seconds for the experimental groups and 579 seconds for the control groups. ANOVA did not indicate a statistically significant difference between the means.

Coding of Backchannel Responses
The video-tapes were made using three high-resolution cameras, two in zoom and the third in normal mode. The two zoom-mode cameras filmed a split screen close-up of images of both participants side by side; the normal-mode camera filmed a full screen of the two participants facing each other. All three screens were synchronized into one picture on the TV screen, with the split screen on top of the full screen. A high-resolution TV/VCR was used to score backchannel response activities.
The operational definition of a backchannel response is any verbal or nonverbal (nod only, not smile) act occurring during the conversation in a non-intrusive manner (not interrupting the speech turn of the current speaker). It can be presented as one word (e.g., yeah) or a statement (e.g., oh, I see) or a question (e.g., is that so). If it is posed as a question, it is judged as a backchannel response if its tone is falling, indicating that an answer from the current speaker is not required. If the question is presented in a rising tone, it is an indication that a response from the current speaker is requested. It is then judged as a question in the front channel, not a backchannel response. Examples of backchannel responses are presented in Appendix A. Three trained research assistants, one English-speaking and two bilinguals, made verbatim transcripts of the videotaped conversations. The transcription technique was identical to that of Beaumont and Cheyne (1998) and Kollock, Blumstein, and Schwartz (1985). Although transcripts were available, coders were required to code backchannel responses from the videotape, using the transcripts as references. Prior to coding the data, coders participated in a training session with the following instructions: (1) read the criteria at least twice; (2) watch the tape while reflecting on coding criteria; (3) code for the first time by watching the tape and listening to the dialogue; (4) code for the second time by focusing on nonverbal activities.
Following the training session, the two coders worked together to decide on the backchannel categories. Once the categories were determined, the two coders independently scored 15% of the data. Inter-coder reliability (Pearson correlation) was 0.87. Differences were reconciled by going through the coding standards and viewing the video tape together. In coding the data, coders were required to mark for overlapping and non-overlapping backchannel responses on the transcripts.

Scoring for Speaker Presentation and Listener Recall
Prior to the scoring, all the video-taped conversations were transcribed verbatim. The answer keys for the two open-ended tests were borrowed from Li (1999aLi ( , 1999b The points for each question were allocated on the basis of the number of information units. The smallest string of words with meaning was the unit of analysis (Bales, 1950;Roter & Hall, 1992). Each unit of information was worth 4 points. Some questions required answers containing one unit of information while others contained several units of information. Thus, some questions were valued at 4 points, while others were as high as 28 points. For example, the answer to question 1 in test 1 "why did the patient come to see you" only contained one unit of information: chest pains. Answers to question 2 in test 2 "what are the possible side effects after you take codeine" contained 7 units of information: drowsiness, nausea, vomiting, constipation, an increase in heart rate, agitation, and respiratory problems, a total of 28 points.
For each question, one or several correct answers were provided in the "Answer Key". If the answer was essentially identical to the key, it was scored as 4. If the answer was very close to the correct answer, it was scored as 3. If the answer was related to the correct answer (e.g., describing but not naming), it was scored as 2.
If the answer was remotely related to the correct answer in that the meaning could be inferred, it was scored as 1.
Blank or wrong answers were scored as 0. A scale of 0-4 allowed the scorer to assign appropriate scores to the range of answers given.
For example, question 7 in test 1 was: what was the main reason the patient went swimming? If the answer was to exercise his/her legs, it was scored as 4. If the answer was essentially "swimming is good for his/her legs", it was scored as 3. If the answer was essentially "to do exercise because the patient has difficulties walking", it was scored as 2. If the answer was essentially "for exercise or to stay fit", it was scored as 1.
To score listener recall, the answer keys were applied to the written tests of the physicians (Dialogue 1) and the patients (Dialogue 2). The same scoring standards were applied to score the speaker presentations based on transcripts of the video-taped conversations.

RQ1:
Were there any significant differences between the means of backchannel responses in the experimental and the control condition, in the doctor and patient roles, and in the Canadian and Chinese participants?
Means of backchannel responses of the two experimental conditions (Chinese doctor/Canadian patient and Canadian doctor/Chinese patient) and the two control conditions (Chinese doctor/Canadian patient and Canadian doctor/Chinese patient) are presented in Table 1 (Task 1) and Table 2 (Task 2).
For Task 1, the means for backchannel responses between the experimental conditions were similar to those in the control conditions for the doctors, (p > 0.05). Patients in the experimental condition made fewer backchannel responses than patients in the control condition. ANOVA indicated that the mean of patient backchannel responses in the Canadian doctor /Chinese patient (control condition) was significantly higher than that of the patient backchannel responses in Chinese Doctor/Canadian patient (p = 0.02; df = 3, 36; experimental condition) and Canadian doctor/Chinese patient conditions (p = 0.03; df = 3, 36; experimental condition). No other significant differences regarding patient backchannel responses were observed.
In terms of role differences, doctors displayed significantly more backchannel responses than patients in Task 1 (M = 21.4, SD = 17.9 vs. M = 9.57, SD = 12.24; F (1, 38) = 15.68, p < 0.001). In Task 1, the doctors were the listeners and patients were the speakers. This finding indicated that the listeners made more backchannel responses than speakers.
In terms of cultural comparison, ANOVA indicated no significant differences between Canadians and Chinese in the means of backchannel responses.
For Task 2 (Table 2), no significant mean differences were found between the experimental and control conditions. In terms of role difference, patients displayed significantly more backchannel responses than doctors (M = 28.27, SD = 18.21 vs. M = 6.24, SD = 6.29; F (1, 38) = 63.24, p < 0.001). In Task 2, the patients were the listeners and doctors were the speakers. This finding indicated that the listeners made more backchannel responses than speakers.
In In terms of the percentage of overlapping backchannel responses, the range was from 36.52 to 50.24 for doctors and 53.12 to 69.23 for patients in Task 1 (see Table 1). No significant role differences (doctors vs. patients) or cultural differences (Canadians vs. Chinese) were found in terms of the percentages of overlapping backchannel responses. The categories of backchannel responses by role (speaker or listener) were presented in Table 3. In performing Task 1, the sum of backchannel responses were 855 and 383 for the listeners and the speakers respectively. For both the listeners and speakers, the highest percentages were "nodding" (53.69% and 79.12% respectively). These two percentages were significantly different (z = 2.41, p = 0.02). The next three highest percentages for the listeners were "okay" (21.99%), "repeat" (7.95%), and "oh" (5.85%). The next two highest percentages for the speakers were "yeah" (6.79%), and "okay" (6.53%).
RQ3: What were the characteristics of backchannel responses in the two cultural groups? Table 4 presents the categories of backchannel responses by cultural groups. In performing Task 1, Canadians and Chinese made a similar amount of backchannel response (646 and 592 respectively). For both the Canadians and Chinese, the highest percentages were "nodding" (57.89% and 65.52% respectively). The next two highest percentages for the Canadians were "okay" (23.53%), and "repeat" (7.89%). The next two highest percentages for the Chinese were "okay" (10.30%) and "uhm" (7.43%).
RQ4: Were backchannel responses correlated with speaker presentation and listener recall scores? Table 5 presents correlations among backchannel responses and speaker presentation and listener recall scores in both tasks. As shown in Table 5, backchannel responses did not have significant correlations with speaker presentation and listener recall scores in either task. However, the correlation between speaker backchannel responses and listener backchannel responses in Task 1 was statistically significant, r (40) = 0.34, p < 0.05 and Task 2, r (40) = 0.30, p < 0.05, indicating that in a dyad, the partners' backchannel response behaviors influenced each other. As well, backchannel responses of listeners (doctor role in Task 1) were significantly correlated with those of the speakers (patient role in Task 2; the same person), r (40) = 0.44, p < 0.01), showing that a person's backchannel response pattern was consistent regardless of the role he or she played. No other statistically significant correlations were found.

RQ5: Did backchannel responses increase the enjoyment of the conversation by participants?
Questionnaire Data. Means scores of the responses by cultural group (Canadians vs. Chinese) to the 7 questions using a Likert scale are presented in Table 6. As English is the first language of the Canadian participants, their English language fluency was not rated. The first question in the questionnaire asked the Chinese participants to rate their own English language fluency and the Canadians to rate the English language fluency of their Chinese partners. As indicated in Table 6, there was a statistically significant difference between the mean ratings by the Chinese and Canadians in terms of the English fluency of the Chinese. Canadians rated the English fluency of their Chinese partners higher than the self-ratings of the Chinese. However, both self-ratings and other-ratings were in the range of 'fluent' to 'very fluent'.
As expected, the Chinese were rated more knowledgeable about the Canadian culture than the Canadians about the Chinese culture, F (1, 78) = 12.11, p = 0.001. Both Chinese and Canadians thought that their partners were reasonably relaxed during the conversations but the Chinese were perceived as less relaxed than the Canadians. Both the Chinese and Canadians had high ratings regarding their enjoyment in the interaction.
Of the 13 questions in the questionnaire, 5 had nominal scales. The percentages of respondents in each category are reported below. About half (47.5%) of the Canadians said that their Chinese partners had some language difficulties while 57.5% of the Chinese thought so. An equal number of Canadians (22.5%) and Chinese (22.5%) thought that both language and cultural difficulties existed when they conversed. When asked what they did to overcome the difficulties, the most frequently used methods by the Canadians and Chinese were: "slowed down" (22.5% vs.17.5%), "repeated the word or sentence" (20.0% vs. 30.0%), "asked questions for my partner to explain" (22.5% vs.17.5%), and 'paraphrased" (17.5% vs. 10.0%). About one-third of the Canadians (35.0%) and the Chinese (30.0%) thought that the Canadians controlled the flow of the conversation, although 27.5% of the Canadians and 40.0% of the Chinese thought that they had equal control of the conversation. The remainder reported that the conversation flowed easily without anyone in control. When asked about their perceived social status, 67.5% of the Canadians and 62.5% of the Chinese reported that they had equal social status. However, 30% of the Chinese thought that their Canadian partners had higher social status than themselves, while 27.5% of the Canadians agreed that they had higher social status than their Chinese partners. The majority of the Chinese (75.0%) reported that the Canadians were linguistically more advantaged than themselves, compared to 52.5% of the Canadians. As indicated in Table 6, the Canadians perceived some communication difficulties in their Chinese partners although they thought their English language was somewhat fluent. The Canadians were aware that the Chinese were not very relaxed although the Chinese thought the Canadians were.
Correlations between frequencies of backchannel responses and variables in the questionnaire were calculated. The only significant correlation was between the frequency of backchannel responses in performing Task 2 and the enjoyment of the conversation (r (40) = -0.20, p < 0.05), an indication that the higher the frequency of backchannel responses, the lower the enjoyment of the conversation.

Discussion
The data generated four intriguing findings; each will be discussed below. First, Chinese participants in patient role made more backchannel responses than Canadians in performing Task 2. This finding is in agreement with previous research (Li, 2006). Comparing frequencies of four dyadic combinations, Li found that when a Chinese talked with another Chinese, the frequency of backchannel responses was highest, followed by the Chinese/Canadian combination, and the Canadian/Canadian combination. What are the possible explanations of these differences? Is it a mere habit? Do Chinese, as a rule, have the tendency to respond through the backchannel more frequently than Canadians? If so, what are the implications for Canadian-Chinese conversation? These issues warrant further research in future studies.
Second, in terms of the types of backchannel responses, the Chinese and Canadians presented similarities as well as differences. They both nodded a great deal and used "okay" quite frequently. The difference between the Canadians and the Chinese lay in the categories of "repeat", "uhm", and "yeah", the former repeating more than the latter, and the latter using more "uhm" and "yeah" than the former. These findings seem to be consistent with previous research (Li, 2006). Li observed that when a Chinese communicated with a Canadian, the Chinese did not use "repeat" whereas the Canadian did. However, when a Chinese talked with another Chinese in Chinese, they used "repeat", and when a Canadian talked with another Canadian, they did not use "repeat". It can be reasoned that the Chinese may not feel comfortable enough to use "repeat" when conversing with a Canadian. This reasoning is supported by the questionnaire data -the Canadians reported that their Chinese partners were not very relaxed. When talking with a Chinese, the Canadian may feel obligated to "repeat" because the Chinese person was experiencing some difficulties, as presented above. On the other hand, when a Canadian talked with another Canadian, there was no need to "repeat".
In light of Li's 2006 findings, the similarities and differences in the use of backchannel responses by Canadians and Chinese observed in the present study have theoretical references for language researchers and practical implications for second language teachers and learners. Theoretically, they provide support for communication accommodation theory in that both groups performed "code switching" to accommodate their partners, a phenomenon termed speech convergence in communication accommodation theory (Cui & Lapadat, 2009;Giles et al., 1973;Giles et al., 1987;Giles & Smith, 1979). In the present study, the Chinese did not use "is that so" (Shi Ma), a backchannel response used in Chinese conversations (Li, 2006), an indication that the Chinese switched linguistic code when conversing with a Canadian, providing further support for communication accommodation theory.
Practically, Chinese second-language learners should be reminded that "repeat" is a linguistic habit (Clark & Wasow, 1998;Schegloff, 1997), and they should not hesitate to use it when conversing in English because various forms of "repeat" facilitate content communication (Aguilera & Li, 2009;Li, 1999b).
Third, we found that backchannel response did not seem to play a role in content transmission by either increasing speaker presentation or listener recall scores. Rather, it is a speech habit, a claim supported by the significant correlations between a person's backchannel response scores in the two tasks whether they be in the role of a doctor or patient. Another interesting finding is the significant correlation between the speaker and listener backchannel response scores in performing both tasks; when one partner was responding through the backchannel, the other reciprocated although listeners did make more backchannel responses than the speakers regardless of role or culture.
Finally, the negative correlation between the frequency of backchannel responses in performing Task 2 and the enjoyment of the conversation is an indication that too many backchannel responses may not be a good thing. Unfortunately, we did not ask participants' opinions on the appropriate amount of backchannel response they prefer. Future research could explore this issue by experimentally manipulating the amount of backchannel response in various conditions. Participants' preferences for the right amount of backchannel response could be found through their verbal reports, facial expressions, tone of voice (e.g., whether high frequency makes them impatient) and whether they use verbal devices to cut the other person short. To elicit the right amount of backchannel response is an important issue in intercultural conversation since too much could indicate "being too eager to please", and too little could mean "being cold, uninterested, or distant".
In conclusion, this study contributes to the field in four ways. First, it documented detailed information on the various categories of backchannel responses in Canadian-Chinese conversations. Second, it provided quantitative analyses of the frequencies of backchannel responses by culture and role. Third, it explored the relationship between backchannel responses and content communication. Finally it examined perceived enjoyment of the conversation by participants as well as its relationship to the frequency of backchannel responses; a significant negative relationship raises the delicate issue of how to balance the appropriate amount of backchannel response in intercultural conversations.