3 Research Methods and Strategies
3.5 Data Analysis
3.5.1 Coding
Coding is an interpretive operation in which the researcher “operates as the central processor of data” (Hadley, 2017, p. 33). To perform this operation, the researcher employs
“reasoning from evidence” (Spradley, 1979, p. 8). Codes are tags which the researcher uses to capture the essence of units of data with the view of analysing data. Segments of data are named, that is succinct descriptive phrases are used to summarize the essence of the segment (Hadley, 2017). Codes “are based on a combination of the researcher’s scholarly knowledge and knowledge of the substantive field under study” (Strauss, 1987, p. 34).
Coding is a layered operation which is present in several actions of the GTM research activity. Coding is one of the basic constitutive elements of data interpretation action, it is inseparable from abductive reasoning, it keeps moving the constant comparison of data and the gradual emergence of concepts (see actions in Figure 4). The three layers of coding are:
• Open coding,
• Focused coding, and
• Theoretical coding.
In what follows, these layers are presented as phases of the coding operation. The phases are presented as separate stages only for explanatory purposes. In the actual coding process, the three phases were mapped on each other on several occasions. At this point, it has to be mentioned that I coded all the data twice. First, I used highlighter pens and sticky notes to code the collected data manually, adopting the traditional method. Then, much later in the research process, I used the Quirkos qualitative data analysis software to operate a second tour of coding. Figure 17 contains an excerpt from a think-aloud protocol analysed with Quirkos. The most important gains of the iterative process of data analysis were the opportunity to reevaluate, reinterpret and expand some codes such as language ability, managing linguistic tasks, and limitations on language use.
3.5.1.1 The Open Coding Phase
Open coding is the initial phase in the coding process: it is set in motion very early, almost simultaneously with data collection (Charmaz, 2006; Strauss, 1987; Strauss & Corbin, 1994). I began the open coding process by treating the data analytically following Charmaz’
(2006) advice to be “open to exploring whatever theoretical possibilities (…) can be discerned in the data” (p. 47). In the open coding phase, the researcher is guided by both the research questions and theoretical sensitivity. Open coding entailed a decomposition of textual data into ideas, situations, and events, which appeared encapsulated in the participants’ utterances.
The units of data were named using labels that captured their essence in the sense of analytic summaries (Urquhart, 2013).
In the open coding phase, I identified pertinent textual data units starting from three questions, of which two are suggested in the disciplinary literature:
Figure 17. Think-Aloud Protocol Section Analysed with Quirkos.
The text sections on the right-hand side are units of analysis. They are numbered consecutively. Different colours are used to code each unit of analysis. The colour pattern on the left-side of each numbered unit of analysis signals the code(s) used to categorize it.
Certain units bear more than one code. This is the case of line 110. Orange signals that L3 English was used in the utterance. Pink signals that L2 Romanian was used in the utterance.
Light blue signals that we deal with a mixed utterance. The coloured bubbles on the left-side area are the codes. The size of the bubbles varies as a function of the number of reiterating patterns identified in the transcribed think aloud protocol.
• “What does the data suggest? Pronounce?” (Charmaz, 2006, p. 47)
• What theoretical category does this specific datum indicate? (Glaser, 1978, p. 64)
• What relevance does the data have to the research questions?
The analysis of the pre-task interviews resulted in the open codes listed in Table 5 and Table 6. The open codes present in Table 5 are related to the concept of language and its uses. The open codes listed in Table 6 relate to the concepts of multilingual self, multilingual community and language spaces.
Table 5. Pre-task Interview Open Codes Relating tot he Concept of Language and Its Uses Pre-task Interview Open Codes Relating to the Concept of Language and Its Uses
Note. Open codes are presented along with the focused codes to wich they were included and with some examples from the transcribed interviews. The names of the participants are reduced to initials. English translation is given in square brackets.
While examining the codes in Table 5 and Table 6, two things have to be kept in mind. First, in the open coding phase the names of codes are provisional. As will be seen later, these labels underwent change either to accommodate new ideas, or to describe more accurately the group of similar patterns they subsumed. Second, there are overlaps between the open and focused codes in the steps of the coding process. This is typical of the development of codes and categories because the codes which acquire substance in earlier phases of the process are taken forward.
Table 6. Pre-Task Interview Open Codes Relating to the Concept of Self, Community and Language Space
Pre-Task Interview Open Codes Relating to the Concept of Self, Community and Language Space
Note. Open codes are presented along with the focused codes to which they were included and with some examples from the transcribed interviews The names of the participants are reduced to initials. An English translation is given in square backets. The words in italics are in vivo codes.
As mentioned in Subsection 3.4.3, describing the logistics of carrying out the pre-task interview, I did not follow the same order of questions in each interview session to let the participants find their own voice as recommended by Charmaz and Belgrave (2012). This step proved later, in the process of analysis and coding, an impediment in identifying similar and different patterns of answers in the textual data. It also slowed down considerably the process of open coding. To amend this process, I decided to have the names of some open codes (and, later focused codes) follow the wording of certain questions from the pre-task interview. This way I was able to track back data easier when I needed to return to the transcribed interviews.
In the case of the think-aloud verbal data, the initial coding process aimed at identifying those utterances which captured meaning making of images, L3 content construction, alternances between languages, difficulties in L3-word retrieval and metacognitive thinking.
Unlike the tables dedicated to the codes pertaining to the data collected in the two interviews, Table 7 does not contain example of utterances for each open code because of the particularities of the think-aloud data. These examples are given below in the body of the text in the section dedicated to the presentation and discussion of the findings connected to the think-aloud activity.
Table 7. Open Codes and Focused Codes from the Think-Aloud Protocol Open Codes and Focused Codes from the Think-Aloud Protocol
The open-coding stage of the data derived from the post-task interviews followed a similar strategy to that of the coding of the data originating from the pre-task interview.
The codes developed during the open-coding stage of the post-task interviews are presented in Table 8.
Table 8. Post-task Interwiev Open Codes and the Focused Codes to Which They Were Included, alond with Examples from the Data
Post-task Interwiev Open Codes and the Focused Codes to Which They Were Included, alond with Examples from the Data
3.5.1.2 Focused Coding
The next phase in interpreting was focused coding. Charmaz (2006) distinguished two types of coding: focused and axial, stating that while focused coding allows the researcher more flexibility, axial coding imposes a “frame” (p. 61) on the interpretations the researcher makes. I employed both types of analytical sense-making in the second stage of data coding because of the advantages they present. The open codes listed in Table 5, Table 6, Table 7 and Table 8 were evaluated in terms of their valence, namely, in terms of their degree of combining power, or relative capacity to form links with each other. In the phase of focused coding, the open codes were elevated to a more abstract level, and they were grouped based on their properties and dimensions, paying attention to their relation to the focus of research inquiry. To understand what the process of elevation consists of, it is useful to use the analogy of chemical elements and their valences. The valences give chemical elements their degree of combining power. The open codes, similarly to chemical elements, have different valences, that is, different strength of combining. Some of them possess a high degree of combining power. These types of codes are the focused codes. In the present study, they were employed in developing relationships and in building the interpretive network. Other codes posses a moderate valence or have no valence. The former codes have a limited usability from the perspective of theory formulation and answering the research questions. The latter types of codes are usually discarded. The focused codes developed in the process of abstracting appear listed in the first column of Tables 5, 6, 7 and 8. The development of focused codes and the interpretive network are the first steps taken towards shaping the substantive theory. They involve identifying relationships between open codes, of the kind described by Spradley (1979, p. 111):
• Strict inclusion X is a kind of Y
• Spatial X is a place in Y, X is a part of Y
• Cause-effect X is a result of Y, X is a cause of Y
• Rationale X is a reason for doing Y
• Location for action X is a place for doing Y
• Function X is used for Y
• Means-end X is a way to do Y
• Sequence X is a step (stage) in Y
• Attribution X is an attribute (characteristic) of Y
To give the reader an example of how focused coding proceeded, I refer to the focused code reliance on the L1 Hungarian, which is present in Table 7 and which was developed during the analysis of the think-aloud protocol analysis. This focused code unites the open codes of identifying the task, setting goals, justifying actions, linguistic assistant,
and creating meaning. Based on Spradley’s (1979) relationships, there are two possible ways to think about the connections between the open codes and the focused code: strict inclusion and function.
Strict inclusion.
Identifying the task, setting goals, justifying actions, being a linguistic assistant, creating meaning ARE KINDS OF reliance on the L1 Hungarian (IN THE PROCESS OF L3 NARRATIVE COMPOSITION).
Function.
Relying on the L1 Hungarian IS USED FOR identifying the task, setting goals, justifying actions, being a linguistic assistant, AND creating meaning (IN THE PROCESS OF L3 NARRATIVE COMPOSITION).
In grouping these open codes, I considered the second relationship as being more expressive of the roles played by L1 Hungarian in the compositional process; thus, the relationship between the open codes and the focused one, in this case is, a functional one.
3.5.1.3 Theoretical Coding
Theory, in the present study, is defined “a coherent description, explanation and representation of observed or experienced phenomena” (Gioia & Pitre, 1990, p. 587). The aim of theories is to “make sense of the observable world by ordering the relationships among elements that constitute the theorist’s focus of attention in the real world” (Dubin, 1969, p. 26).
Theoretical coding, termed also “theory condensation” (Goldkuhl & Cronholm, 2010, p. 196) is the most intellectually challenging and taxing phase of the GTM analysis. This is a concluding phase, in which data analysis is moved in a theoretical direction by aggregating the focused codes, frequently referred to as categories (Charmaz, 2006; Goldkuhl &
Cronholm, 2010; Urquhart, 2013). Aggregation refers to the development of relationships between categories, their subcategories, identifying their relational properties and fitting them into a systematic framework which can act as a set of principles explaining particular phenomena. Theory building corresponds to what Strauss and Corbin (1998) termed
“selective coding” (p. 69), highlighting the act of choosing from among the core categories those that have special relevance for the development of the substantive theory, seen as an”
explanatory scheme” (p. 229). In addition to Spradley’s (1979) set of relationship propositions, Strauss and Corbin’s (1998) formal procedure of relating categories was an especially useful tool in theoretical coding. Their category-organizing scheme rests on three straightforward steps, guided by questions where-what-with what consequences, and follows the principle of contextuality (Gergen, 1982).
The triad’s first component—the where step—requires the examination of the broader conditions and circumstances in which the studied phenomenon exists. This examination situates the phenomenon in space and time. The triad’s what step entails the examination of the participants’ actions and responses, both typical and atypical, to issues and situations. In examining the participants’ actions, the researcher needs to look for answers to the question what. The triad’s third step requires the examination of the consequences of the participants’
actions. It seeks the answer to the question of what are the consequences of the participants’
actions, or what happens because of the actions the participants take. Thus, in a nutshell, Strauss and Corbin’s (1998) organizing frame looks at the participants’ actions (second component) in their context (first component) and examines their consequences (third component). By considering the larger context, the organizing frame helps in identifying the category’s dimensions and properties (Strauss & Corbin, 1998).
To illustrate how categories and the relationships between them were identified and developed starting with the open coding through the theoretical coding phase, I describe the way the core categories languages used together and language use demarcated by place/activity took shape. In the open coding phase of the pre-task interview data, I coded demarcated language use those account sections which described how the use of certain languages was well-delimited to particular activities and contexts in the participants’ life. Text sections mentioning how L2 Romanian often acted as a helper in the case of the other languages the participants knew were coded foreign language through the L2. The text sections in which participants mentioned L2 Romanian as a support for the languages they were learning were coded L2-link to foreign languages. In the focused coding phase, the code demarcated language use became expanded into the category language use demarcated by place/activity because, by examining the transcripts, it became clear that the participants’ reliance on their different languages was influenced by the dimensions of physical environment and the actions they were performing. Subsequently, the codes foreign language through L2 and L2-link to foreign languages were incorporated in the category languages used together.
Later in the analysis, during the open coding of the think-aloud protocols, there were a considerable number of the instances in which the participants employed L2 Romanian lexical units to help in the retrieval of L3 English lexical units, resorted to the Romanian-English bilingual dictionary to look up L3 Romanian-English words and several instances in which the participants started to look up an L3 English word in the Hungarian-English bilingual dictionary, then abandoned it to turn to the Romanian-English dictionary. These events indicated that the role of L2 Romanian fitted into the above two categories: language use demarcated by place/activity because the L2 was employed in the activity to look up L3 words (instead of the L1 Hungarian) and languages used together because the L2 Romanian
was employed in connection with the L3 English. Thus, data from the think aloud protocols solidified the ground for the two categories language use demarcated by place/activity and languages used together.
The open coding of the post-task interview data further substantiated and added nuances to the category of languages used together in terms of the open code convenience-guided use of the L2 Romanian, where convenience was defined as “comfort and ease”. This open code was grounded in those accounts in which participants explained their use of the L2 Romanian words as springboards in the retrieval of the L3 English lexical units because this was easy and at hand (in vivo code).
In the data collected during the post-task interviews, an interesting aspect of dictionary-use became visible in the case of the participants who attended the Hungarian minority education. Although English was taught at school through L1 Hungarian, the participants reported that they used the Romanian-English and English-Romanian bilingual dictionaries to familiarize themselves with the requirements of the exams. I coded these accounts first using the L2 Romanian-L3 English dictionary in the L1 Hungaian education in learning the L3 English. Then, this open code was connected with the languages used together category based on a strict inclusion relationship (Spradley, 1979). In the phase of theoretical coding, both using the L2 Romanian-L3 English dictionary in the L1 Hungaian education in learning the L3 English and convenience-guided use of the L2 Romanian acted as dimensions of the languages used together category. At the same time, it is easy to discover a second relationship between the open code using the L2 Romanian-L3 English dictionary in the L1 Hungaian education in learning the L3 English and the category language use demarcated by place/activity if we interpret the use of the L2 dictionaries to look up the L3 English words as demarcated by the necessity to comply with the educational requirements. Based on Spradley (1979), the relationship is a causal one.
Figure 18. Categories and Their Relationships to the Core Categories in the Phase of Theoretical Coding.
Round-cornered rectangles are categories. Square-cornered rectangles are core categories.
The arrows symbolize the relationships between the categories and core categories.
In the theoretical coding phase, my analytical reflection was supported by Strauss and Corbin’s (1998) organizational framework. The integrative diagram in Figure 18 shows the two categories language use demarcated by place/activity, and languages used together and their dimensions, convenience-guided use of the L2 Romanian and using the L2 Romanian-L3 English dictionary in the L1 Hungaian education in learning the L3 English.
Based on Strauss and Corbin’s (1998) category-organizing elements, I developed the following relationships between the categories:
1. Based on the condition/circumstances element (Where?): The participants used their L2 Romanian and L3 English together mainly in educational contexts.
2. Based on the action element (What?): The participants for whom L2 Romanian was the language of education used L2 Romanian for a series of school-activities, among them they were learning L3 English through L2 Romanian, and relied on the Romanian-English, English-Romanian dictionaries for identifying word-meaning.
The participants for whom L1 Hungarian was the language of education were learning L3 English through L1 Hungarian; however, they were using the Romanian-English, English-Romanian bilingual dictionaries for identifying word-meaning. Each language within the triad: L1 Hungarian-L2 Romanian-L3 English was used in the given educational contexts depending on the place and the activity spheres.
3. Based on the consequences of the participants’ actions (What are the consequences?): The participants’ intensive and extensive reliance on the L2 Romanian in relation to their L3 English, made them consider L2 Romanian as a gateway or facilitator for L3 English (and the other foreign languages).
Consideration of the question as to why long-term use of languages together may contribute to a preference of relying on one of the languages as a support for the other in situations of word-retrieval difficulties will be taken up in connection with theory formulation, in Section 4 and Section 5.