The Efficacy of Unguided Conversation on Fluency and Speaking Confidence: A Case Study in South Korea

To date, there is limited research into the measurable benefits of unguided and unstructured conversation practice (typically a popular, in-demand service) for adult English language learners. This project looks to determine the differences in measured oral fluency as well as self-reported speaking confidence of South Korean university students before and after several months of conversational English practice with native English instructors. Participants were given weekly one-on-one "free talk" sessions with no grammatical instruction and very limited corrective feedback for a total of one hour. Sessions were recorded at early and late program intervals to calculate the average length of runs and speech rate. Moreover, students completed preand post-program surveys to estimate their overall confidence levels in L2 speaking. This project found limited change in fluency without direct corrective feedback. Furthermore, our measurements of students’ self-reported speaking confidence revealed inconsistent trends from the first to last sessions. This research seems to indicate that loosely structured, “coffee shop” style lessons should adopt a more structured curriculum in order to see more pronounced and measurable growth over time.


Introduction
A considerable amount of second language learning is done orally, with emphasis on natural language input and output high on student and instructor goals. It is also common to see advertisements for students or instructors looking for strictly conversational classes, meeting in coffee shops or online, and focusing purely on speaking with unguided free talking sessions. Further, conversation exchange groups and partners, or tandem language learning, are seen as a relatively inexpensive way to improve one's language skills and have become tremendously popular. Of course, while these groups may be purely conversational, peer language feedback is often used to correct errors, though this primarily takes place through unstructured conversations and free talking (Acar & Kobayashi, 2011). Clearly, oral output is a viable aspect of most language learning programs, but the extent to which output alone facilitates language acquisition remains debatable, with some research suggesting that a certain level of exposure to relevant input is necessary to promote noticing during language acquisition (Izumi & Bigelow, 2000).
Of course, prior research assumes some degree of formal guidance throughout the input and output periods, but there is limited research into the effect of unguided input and output over time. The Output Hypothesis developed by Swain argues that learners will navigate syntactic processing during speech production, particularly when pushed to produce appropriate responses in social situations, and that language production enhances both fluency and accuracy (Abdi et al., 2012;Swain, 1985Swain, , 1995. This research assumes that students will be producing language under the constraints of a learning environment with some attention to linguistic form being prevalent. However, this is not the case in more free flowing settings that are becoming popular with both students and instructors. Many general L2 courses will involve a wide range of lessons looking to target learners' input. However, some research has focused on the effects of 'free talk' activities on language acquisition. A study by Gorkaltseva et al. (2015) including 24 students in general English courses at a University in Russia focused entirely on the improvement of oral fluency using the cognitive-communicative approach, and utilized regular free-talking sessions aimed at student interests, such as sports and weekend activities. The instructors moderated conversations, and tried to limit corrections to repetitive issues in participant speech. This study revealed "considerable progress in enhancing oral fluency" (Gorkaltseva et al., 2015, p.146). This study is different in that the conversations were in groups, and were not strictly unguided, but allowed for instructor corrections. Considering the current prevalence of students engaging in mostly unguided conversational practice, it is imperative to see what practical benefits might be derived over time.

Defining fluency
Fluency is a commonly used term in foreign language teaching and is often the stated goal for language students; however, as a concept, it is vaguely understood and challenging to define. Fluency is often used to define oral proficiency, and fluent in describing a person who can use a language, specifically a foreign language, effectively and with ease (Baker-Smemoe et al., 2014;Chambers, 1997) Many instructors and students feel that they can perceive fluency in others naturally but might have difficulty describing what aspects of a person's speech create the perceived fluidity. Despite the many learners whose stated goal is to become fluent, a widely agreed-upon definition remains elusive. Therefore, researchers need to continue working on how the concept of fluency can be accurately measured (Kormos & Dénes, 2004).

Measuring fluency
The difficulty arises when determining quantitative measurements for a vaguely defined concept. Some studies have used educated raters who listen to and score a speaker's fluency based on a set of criteria determined by the researchers (Derwing et al., 2007). This could lead to bias or variation among raters, and more objective measurements are common. Other research has looked at disfluencies in learner speech, such as reformulations, replacements, false starts, repetitions, hesitations, and pauses or silences (Abdiet al., 2012;Foster & Skehan, 1996;Kormos & Dénes, 2004). Further research has used temporal measurements in recorded speech, which include speech rate, average length of runs, length of time on tasks, articulation rate, mean length of pauses, silent pauses per minute, number of filled pauses, and stressed words per minute (Chamber, 1997;Foster & Skehan, 1996;Hieke, 1985;Kormos & Dénes, 2004;Préfontaine et al., 2015;Tavakoli & Skehan, 2005;Yuan & Ellis, 2003).
Our study employed temporal forms of measurement, which have been shown to accurately measure fluency growth in previous studies. First, we measured speech rate as the number of syllables per second. Second, we measured the mean length of run as the average length of speech in syllables between pauses of .25 second or longer. The nature of our study revolves not around predetermined speaking tasks but instead natural conversation patterns, which includes negotiation of turn-taking on the part of the speaker. Therefore, we felt that these metrics would most accurately measure fluency growth in the participants.

Speaking confidence and language anxiety
One aspect of proficiency with any learned skill is achieving the confidence to perform. Language learning is no different, and there are arguments discussing the factors that possibly affect speakers' confidence levels and proficiency. According to Krashen (1982), reduced anxiety is one of the factors that directly relates to success in second language proficiency. It has been estimated that up to a third of language learners suffer from some degree of language anxiety (Worde, 1998;Zhang, 2011) and that students feel considerable more anxiety in language courses (Cutrone, 2009). Language learner confidence could be one factor that prevents higher levels of English proficiency (Aida, 1994), and it has been seen where a lack of confidence can cause students to fear communication, resulting in students refusing to participate in situations where they could potentially improve their language abilities (Daly, 1991;Khusnia, 2016).
Concerns and anxieties are, of course, situational, and therefore speaking fluency could vary within individual learners due to momentary feelings of anxiety or pressure (MacIntyre & Gardner, 1994;Préfontaine et al., 2015;Wood, 2010). Some suggest that communication apprehension, or an anticipated fear of speaking (Cutrone, 2009;Horwitz et al., 1986), could be mitigated with activities focused on oral output in controlled settings. Further, some researchers argue that perceived language proficiency causes doubt, while others have shown self-confidence and feelings of inferiority cause language anxiety (Liu & Jackson, 2008;Brown, 2008).
The impact of corrective feedback has been connected to speakers' anxiety revealing that learners may be reluctant to participate due to negative assessments (Cutrone, 2009;Krashen, 1985). The present study will test if unguided speaking sessions, or speaking sessions with no corrective feedback or instruction, in a somewhat informal environment could improve a student's confidence and reduce language anxiety over time, which would constitute one practical benefit of this style of language learning.

Overview of study
This case study looks at six Korean university students with self-assessed intermediate English abilities and a desire to improve their conversational abilities. Students receive weekly unguided conversational sessions with limited corrective feedback, and their fluency and confidence levels are measured at the beginning and end of the project using speech rate and mean length of run as defined previously. Relevant sessions are recorded and transcribed, with pauses measured to determine the mean length of run and syllables per second calculated to determine speech rate. The goal is to measure changes in English fluency over the course of the project.
Further, students are given questionnaires at the beginning and end of this project (see Appendix A). One section of this survey is constructed to determine the participants' perceived confidence level when speaking English. These responses assist in determining whether participants show any growth in L2 speaking confidence.

Research questions
Research into the measurable improvements made by students working purely with unguided conversational classes is currently limited. This case study looks to determine what changes can be seen in learners' fluency and confidence after consistent conversational practice and seeks to answer two questions:

1.
Do students show measurable improvement in their fluency when given consistent opportunities for unguided conversational practice? 2.
Do students perceive an improvement in their L2 speaking confidence after receiving consistent opportunities for unguided conversational practice?

Participants
The participants were six Korean university students (four graduate and two undergraduate) enrolled at << masked for blind review>> in South Korea in 2020. Students who attend this university are required to submit a proficiency score in one of several official English proficiency tests, such as TOEIC, TEPS, or IELTS, and participate in an English interview, thus ensuring there would be no participants with too low a conversational ability. The ages of participants ranged between 18-21 and 27-31 for undergraduate and graduate students, respectively. The group consisted of two female and four male students.
This study employed purposeful, non-probability sampling by selecting participants among volunteers who met the following criteria: the self-diagnosed need for improvement in speaking/listening ability, and availability to meet for one hour per week, either twice for 30 minutes or once for 1 hour. The participants were selected after a survey was emailed to the student population inviting those interested in one-on-one conversational English sessions. Fifty-four students indicated interest in participation; out of participants who indicated an intermediate English level, six participants were chosen based on a desire for improved conversational English ability and availability for this project.
Cannot "talk freely with foreigners."

Self-reported speaking confidence
The pre-program survey gathered information on the participants' background, ability to participate in the study, general English experience, perceived English ability, and 12 questions regarding speaking confidence (see Appendix A, adapted from Griffee (1997). These questions were used to establish a baseline of participants' self-assessed English confidence and ability before initial meetings with native English instructors. Questions regarding general English experience and perceived English ability were a combination of multiple-choice and short-answer. The sections pertaining to self-reported speaking confidence used 5-point Likert scale questions, ranging from "Strongly Agree" to "Strongly Disagree." The confidence questions were based upon two of the three factors affecting "confidence" in Griffee's (1997) research: English ability (proper use of grammar, pronunciation and vocab) and willing engagement (feeling positive emotions and ease when speaking with native English speakers). A follow-up survey was administered at the end of the study, which repeated the questions on speaking confidence to assess any changes quantitatively.

Oral fluency
After completing the pre-program survey, the six participants met with five native English instructors for a total of twelve hours divided into hourly or thirty-minute sessions over twelve weeks through one-on-one sessions using Zoom, an online video communication tool. Participants rotated instructors after four to six hours to avoid familiarity affecting fluency. To encourage and support a natural conversation environment, the sessions were unstructured or loosely structured natural, conversation-focused meetings. Instructors were permitted to speak slower, discuss cultural differences, and address communication breakdowns in a natural manner. However, explicit feedback regarding language errors, such as offering metalinguistic clues or reformulating speaking errors, was expressly avoided.
During the first session, participants were asked four out of eight possible open-ended questions and asked the remaining four questions in the last meeting (see Appendix B). The eight questions were chosen from a bank of conversation-friendly questions related to personal preferences and were meant to allow students to freely answer and demonstrate fluency. These questions helped standardize the conversations between all participants to simplify the analysis by allowing participants to elaborate on the question topics for longer speech segments. The recordings from the first and final sessions were divided and transcribed by the instructors in the following way.
The recordings of the responses to the eight questions were analysed through Audacity, a program in which we were able to measure the length of audio snippets in seconds/hundredths. The participants' answers to the eight questions were pruned since the responses were not from timed speaking tasks but rather taken out of the natural conversation. In other words, standalone fillers such as "uh" and "um" did not count toward the number of syllables in the utterance, but they did count towards the length of the utterance in seconds/hundredths. Pauses of over .25 second were considered the minimum separation between utterances, and the length of the utterances was rounded to the nearest .25 second (Kormos & Dénes, 2004). The analysis involved counting the total number of utterances in the recorded selection, the length of each utterance in hundredths of a second, and the average length of all utterances. Repeated words (e.g., "I went to… to the concert") were included for time measurements but not included for syllable counts. Then, the speech rate (number of syllables per second) and mean length of run (average length of speech between pauses of .25 seconds or longer) were compared between each participant's first meeting and last meeting.

Results
To review, this case study attempted to measure changes in oral fluency and speaking confidence of six participants before and after 12 hours of unstructured conversation practice.
Our quantitative fluency measures were speech rate and mean length of run. These values were derived by analysing segments from each participant's first and last conversation session in which participants responded to eight questions. Our qualitative confidence values were measured by comparing opening and closing survey results of twelve questions modelled after Griffee's (1997) survey. These 5-point Likert-scale survey questions could be answered in a range where a "Strongly Agree" corresponded to a numerical value of 1 and a "Strongly Disagree" corresponded to a value of 5.

Research question 1: Do students show measurable improvement in their oral fluency when given consistent opportunities for unguided conversational practice?
Overall, measures of our participants' fluency showed negligible change (see Table 3). Changes in both the speech rate and mean length of run averaged less than one full syllable between the first and last sessions. For most of our participants, there was very little change in speech rate, with increases occurring in only two out of six participants. Mean length of run only increased for one participant. For both metrics, increases did not exceed more than one syllable of difference. A majority of participants' speech rate decreased slightly by the end of the program, and their mean length of run showed a greater decrease. At the end of the unguided speaking sessions, the students' responses in the post-program survey were generally similar to their initial survey responses. With regards to the specific questions regarding speaking confidence, six out of twelve confidence questions showed a slight positive change. The greatest positive changes occurred in the following two confidence questions in the survey: "I can show an English-speaking visitor around the campus and answer questions," and "I am relaxed when speaking English". There were smaller net positive changes in the six students for the following questions: Generally, in contrast to the positive differences, there were no outlying individual student responses that showed a larger negative change in confidence to any particular question than others.
When comparing students' confidence survey responses overall, only two out of the six students showed a large difference in values between the initial and follow-up surveys. However, these two students had divergent results. The student with a lower overall starting confidence scored two or more points higher by the end of the program in the following questions: o I can be interviewed in English. o I can show an English-speaking visitor around the campus and answer questions. o I am relaxed when speaking English The student with higher starting confidence scored lowest, all by the same decrease of 2 points, in the following questions: o I can be interviewed in English. o I would like to study in an English-speaking country. o I like speaking English o I can converse in English with native speakers. o I look for opportunities to speak English.
Overall, we could not derive any improvements across the student responses in either speaking confidence or oral fluency.

Discussion
This study allowed participants to have consistent opportunities to speak freely with native English instructors in a relaxed environment over a twelve-week period. As mentioned previously, few studies have looked at measurable benefits of unguided conversational practice over a set period of time. While we recognize that this type of learning is becoming more common, it is important for language teaching professionals to review the efficacy of such learning strategies. Of course, the present work does not seek to disprove claims that language output is necessary in order to improve upon both fluency and accuracy in students' overall language production, as is supported by other research (Abdi et al., 2012;Swain 1985Swain , 1995.
However, when focusing purely on unguided language production, students saw mixed results in terms of measurable indicators of success.
Our first research question asked whether or not students show measurable improvement in their oral fluency when given consistent opportunities for unguided conversational practice. The fluency measurements (speech rate and mean length of run) taken in the final recorded conversations with our subjects failed to show significant improvements over the initial recorded conversations, despite the significant time spent receiving one-on-one conversational practice with native English instructors. Instead, four out of six participants saw a slight decline in their speech rate and in their mean length of run. This suggests that our participants saw no measurable improvement in their overall fluency through this project.
Further, our survey covering students' speaking confidence showed minimal change from the start to the end of the experiment. This is in reference to our second research question, which asked whether students perceive to have improved their L2 speaking confidence after receiving consistent opportunities for unguided conversational practice. Low language confidence and speaking anxiety have been shown to negatively affect language learners' overall performance (Brown, 2008;Liu & Jackson, 2008), as well as hinder learners from fully participating in activities that could further improve their language abilities (Khusnia, 2016). Therefore, measuring participants' perceived speaking confidence before and after weekly interaction with native English instructors was an important goal of this project. Based on our survey results, only half of the confidence-based questions showed any positive change, and this change was minimal among participants. Additionally, several confidence-based questions revealed a small net negative change from the initial to the final survey. Overall, no evidence was seen to suggest that participants felt a measurable increase in their overall confidence when speaking English. As a group, the participants' perceived improvement in speaking confidence during this project was negligible.
However, there were two outliers in the Likert-based confidence questions, with one student showing a significant increase in confidence and another showing a sharp decrease. Due to our relatively limited number of participants, we believe a more consistent change would need to be present across the participants in order to indicate growth in confidence by the end of the experiment. Of course, this was a preliminary project with several limitations that, if addressed, could produce more diverse findings in future studies.

Limitations
Despite time with native English instructors in a consistent environment, participants as a whole saw little or no measurable growth in oral fluency or speaking confidence. The ability to extrapolate these results is limited due to our study's small number of participants. Further research is needed with a larger number of participants, a more longitudinal investigation, and different fluency measurements to obtain more significant statistical data. Specifically, future work should assess how continued unstructured conversations may affect factors outside of the scope of the measurements within the current study.
After the study, there was a case of one participant who expressed improvements on an international standardized test due to weekly conversation practice with an instructor. This implies that there may have been a lack of self-awareness of changes in English fluency among participants at the time of the exit questionnaire. There is a possibility that subconscious improvements were not recorded but later realized only after the student participated in an evaluation or activity requiring English output in a different context. A follow-up survey could reveal further changes in English fluency factors that were not immediately noticed by the participants at the end of the study. Also, certain improvements may not have been evident quantitatively and needed a qualitative method of assessment. Thus, future studies could explore how confidence and comfort while speaking in English as a second language may change with extended exposure and experience conversing with a native English instructor with the inclusion of follow-up surveys and interviews.
Not only was there a limitation in the number of participants and assessment methods, but the analyzed audio segments may not have offered authentic measurements. This could be because the fluency analyses were limited to the oral responses to a list of target questions and not other topics or parts of the conversations. Moreover, the instructors interviewing the participants also analysed the data sets individually at different rates, which may have impacted unbiased results. However, time and resources were limited.

Conclusion
The present study looked at changes in English oral fluency and speaking confidence in a limited number of Korean undergraduate and graduate students after consistent yet unstructured weekly conversations with English native instructors who gave little-to-no corrective feedback. It is important to emphasize that these results do not suggest that unstructured conversations have no value or effect on second language fluency. As previously stated, there may be benefits of free-talking that were not assessed in this study. Bower and Kawaguchi (2011) found negotiations of meaning occurred more often than feedback in language exchanges, which is another important aspect of second language learning. The present study also found that although instructors avoided feedback, negotiations of meaning were necessary for natural conversation to continue communication. However, language educators must be aware that conversation-focused exchanges may not be the most effective method of fluency improvement.
Language exchange programs are expected to engage in passive or unfocused language practice with unstructured conversations and free talking (Acar & Kobayashi, 2011). Based on the results of the current project, we believe that unguided conversations with teachers and educators should be used in conjunction with other pedagogical activities to see pronounced growth in students' output abilities and overall confidence. While free talking and language exchange style learning can have practical benefits, these strategies should be accompanied by a more scaffolded and goal-based approach to language acquisition. Such conversational programs may be more effective by adding a structured component, curriculum, and/or clear corrective feedback.
Overall, as location-independent education grows, online one-on-one language instruction will become more commonplace. According to our findings, language educators may be more effective when offering specific feedback and language-focused guidance rather than having unstructured, informal conversations. Therefore, it is important to consider the most effective structure that we can offer to assist our students' language progress. □ No □ Maybe 9. These sessions will only be free conversation, without feedback, with and English speaker. Is that okay? □ Yes □ No 10. How many years have you studies English? □ 0-2 years □ 3-5 years □ 6-8 years □ 9-10 years □ 11-12 years □ 13-14 years □ 15+ years 11. What kind of activities do you do to improve your English?
12. How many hours do you spend improving your English in a typical week? □ 0 hours □ 1-2 hours □ 3-4 hours □ 5-6 hours □ 7+ hours 13. Check any of these that you have done. Her research interests include English language teaching (EFL), discourse analysis, and second language acquisition. (email: jmanning@kentech.ac.kr) Natasha Powell instructs students at the Pohang University of Science and Technology. She has an engineering/design background and a constant interest in Design and the Design process. As a language and technical writing instructor for nearly a decade, she has been merging her interests to find ways to enhance education through the design thinking process. (email: npowell@postech.ac.kr )