The current study was composed by a survey study and a multiple case study. As a result, the data set had two main components, survey responses from the sixty-five respondents and the data collected through the multiple case study (Merriam, 2009) involving five subjects. The case study research methodology was chosen because, firstly, it is believed to be able to "provide insight into an issue or refinement of theory” (Stake, 1998, p. 88), and it is most appropriate for studies which are descriptive, dynamic, rely upon naturally occurring data (Braine, 2002; Petrić & Czárl, 2003), and aim to reveal complexities and intricate interaction between multi-dimensional factors in a given context, which are endemic to situated
qualitative research (Atkinson, 2005). Secondly, Addison (2007) believed that, by putting the subject into a specific cultural, social, and historical moment, case study was a good fit for describing and critically understanding the situated nature of human experience such as literacy. Thirdly, Yamagata-Lynch (2010) argued that activity system analysis is compatible with case study research, and doing case study within clear and bounded systems in natural settings helps to make data collection and analysis more organized and focused. Fourthly, one of the goals of conducting this research was to "dismantle stereotypes of cultural patterns in writing and of writers labeled simplistically as representatives of their respective cultures" (Casanave, 2005, p. 29). Accordingly, the research design of the current study was not only to look at similarities among first-year international Chinese undergraduate students in their
academic writing, but also more importantly, to observe particularities and appreciate the idiosyncrasies that characterize human experience. To this end, the multiple case study design is a viable choice.
Data analyses were thus performed using a two-step procedure. First, a quantitative analysis was conducted on the survey data. The main objective of this analysis was to identify basic trends and describe the general situation. Second, a multiple case study was conducted on the five students' data. In consideration of the merits and shortcomings of different research techniques, a combination of multiple methods was adopted. Surveys, real- time screen recording observations, interviews, as well as instructors' feedback were used to examine the research topic from a number of different perspectives and achieve as much triangulation as possible. The combination of methods is similar to Perrin's (2003)
"progression analysis" method in combining objective with subjective data to capture what the participants are actually doing and how they account for it.
To be more specific, survey was used to collect responses from a bigger group of
students in the first part of the study for gaining a general picture of the situation in question and identifying potential case study participants. Survey is no doubt convenient and effective in collecting data for analysis about general trends and preferences, but it tends to impose preconceived classifications and elicit decontextualized responses (Braine, 2002). As a result, the survey data needs to be interpreted with caution.
In the multiple case study section, data analysis was carried out on participants' survey responses, their screen recordings, semi-structured interview responses, and the instructors' feedback on their writing assignments. Real-time screen recording allows researchers to observe what happened on the screen during students' writing process and obtain more detailed information about local, contingent factors that might influence particular cases of writing and writers' actions. In addition, it offers a less obtrusive way to glimpse how participants handle each observable challenge along the way and how they select, use, struggle, and proceed with the use of web-enabled resources.
Real-time screen recording has gained increased popularity as a research tool over the past two decades for recording and analyzing the richness in detail of online writing processes. Park and Kinginger (2010) used screen recordings, together with retrospective reflections and corpus search queries, for an in-depth, moment-by-moment analysis of an L2 writer’s composing process. Geisler and Slattery (2007) recommended the use of video screen capture to make visible phenomena that might otherwise have gone unnoticed in digital writing. According to them, the screen recordings can be analyzed at different scales depending on the phenomenon of interest in the study. In their own study based on the activity theory framework, Geisler and Slattery analyzed their screening recordings in a two- stage process beginning with what can be directly observed from one frame and then moving to what requires inferences across several frames. The data analysis of the screen recordings
collected from the multiple case study subjects in the current study involved a combination of both the directly observable phenomena and those that required inferences and calculation.
However, screen recording alone does not provide answers to all the research questions; therefore, guided by the theoretical framework and the research questions, a semi-structured (Spradley, 1979) and open-ended (Patton, 2002) interview was used to gain complementary information and have participants reflect on that particular writing experience and on their writing experiences in general. The specific approach used for interviews in this study was what Cohen, Manion, and Morrison (2000) referred to as an interview guide approach, where the topics and issues are specified in advance as an outline (see Appendix B) and the
interviewer decides the sequence and the emphasis of the questions. Such a method "offers a systematic way to cover salient issues, yet is flexible enough to allow for follow-up of interesting possibilities when participants introduce their own ideas and connections" (Hyland, 2005, p.185). However, like many other research techniques, interview is not without its problems. The most notable one may be the double (and often conflicting) functionality of interviews (Hammersley & Atkinson, 1995), which refers to both reporting what happens (etic description) and what the participant thinks about what happens (emic description).
Despite its potential drawbacks, the interview method may help to enhance learning, as reflection is "an engine for the reconsideration of knowledge" (Kurek & Hauck, 2014, p.135).
Flinn (1987) suggested that when students become more aware of their writing decisions through the help of reflection in the retrospective interview, they can learn to manage their own writing processes more effectively. While the recounts or perceptions provided by the participants in interviews do not necessarily match with the reality and are sometimes not accurate (Park & Kinginger, 2010), it provides an opportunity for researchers to obtain a more comprehensive understanding of the phenomenon under study and for interviewees to reflect on their literacies practices.
Furthermore, in order to lower the participants' level of anxiety and make sure they fully understood the questions and expressed themselves, the interlocutor usually asked the questions in both English and Chinese, and the interviewees were allowed to answer the questions in whatever language(s) they preferred. Although the students ended up using Chinese most of the time during the interview, they did often code-switch between English and Chinese to help clarify their ideas.
During the data analysis process, the CHAT approach and the multiliteracies theory perspective were used to explore the goal-driven and artifacts-mediated human behavior of the case-study subjects during their academic writing process and to investigate
contradictions and tensions emerging from the analysis. In addition, according to Hyland (2005), in composition studies, "from a social perspective, we must also make decisions about how these perceptions and beliefs carry traces of wider participation frameworks not
immediately accessible in the composing context" (p.187). As a result, analysis will always involve inference and is always selective. This is exactly what happened in the current study. Although multiple layers of data were collected and analyzed in this study, the analysis was tentative and by no means exhaustive. However, by grounding the discussion in triangulation of several data sources, effort was taken to make sure that the interpretations are not pure speculation either.