Thesis (Ph.D.) - Indiana University, School of Education, 2019The present study investigated The Mixxer language exchange site (https://www.language-exchanges.org/) as a typical informal online community of self-directed language learners with a focus on how learners interact with each other. Learners’ perception was also studied regarding how their language learning was impacted by their participation in the community. An exploratory sequential mixed methods design was adopted for this work. Data collection started with the iterative feature analysis of The Mixxer website and document analysis of website posts and records, along with an interview with the website administrator. Eight English learners and eight native speakers of English were then interviewed regarding their language learning experiences at The Mixxer after an initial understanding of The Mixxer environment was obtained. Data from interviews with learners were analyzed thematically to inform the development of the survey questionnaire for the last stage of data collection. During the last phase, a survey was distributed to over 12,000 users of The Mixxer. Importantly, more than 500 total responses were received. Of that total, 396 responses were analyzed in SPSS and MS Excel to reveal the interaction patterns and related experiences of all learners at The Mixxer. Analysis results showed that learners conducted conversational language exchange and writing corrections at The Mixxer. The user profile and search system as well as the messaging/chat tool of The Mixxer facilitated the connection of language partners. Learners used Skype or other communication tools to meet with their partners. Meanwhile, a smaller group of learners engaged in writing correction activities in the forum-style Writing section of The vii Mixxer. The Mixxer learners had one to four language partners on average. Among the 71.21% learners who had language exchange experiences at The Mixxer, they typically had one to four 31- to 60-minute exchanges each month. Ordinary learners attempted to arrange equal language practice opportunities for both parties during the exchange. They liked to chat casually with partners, ask each other questions, and teach basic knowledge of the language. They also corrected each other’s mistakes or gave other types of feedback. Some other learners had more structured activities for their language exchange engagement, such as reviewing different types of learning materials together or designing meaningful activities for language practice. Learner’s preferences of language partners were also discussed. Analysis of qualitative and quantitative data with the Community of Inquiry (CoI) framework suggested a strong social presence in The Mixxer community with a variety of indicators and examples. Cognitive presence was identified from learners’ regular language input and output practices during the exchange. Teaching presence in The Mixxer community was largely performed by learners themselves in the categories of design and organization and direct instruction. Evidence was also found for learning presence that effectively extended the CoI framework suggested by Shea and Bidjerano (2010, 2012). In addition, an inspection of learners’ perceived impact of their Mixxer experience, reflected many benefits of conducting language exchange in this community, such as the opportunities to practice the language with native speakers, improvement of language skills, and motivation for continued language learning. A series of challenges were also reported by The Mixxer learners, including time-scheduling issues, unstable Internet connections, difficulty in finding a partner, and so forth. Other language learning efforts were also studied for a better understanding of the role that participation in The Mixxer community played in learners’ entire language learning endeavors. The interactive relationships between learning context, learners’ self-directed learning (SDL) attributes, and SDL processes were exemplified in the present study. Improvements to Song and Hill’s (2007) SDL model were proposed by incorporating CoI components.