Theoretical Sampling, Constant Comparison of Data and Theoretically

Theoretical sampling is one of the special actions of the GTM and refers to sampling aimed at substantive theory construction (Glaser & Strauss, 1967). Its essence lies in the aim of the researcher to track down and collect data about qualities, activities or events characteristic to those individuals who are in the possession of the information and knowledge that can contribute to answering of the research questions (Charmaz, 2006;

Glaser, 1998; Morse, 2007; Strauss, 1987; Urquhart, 2013). In the case of the present work, theoretical sampling meant contacting multilingual individuals from whom I could collect information on whether and how they used in terms of lexical retrieval the languages they knew when they had to come up with a story in a non-native language, a story which was based on a series of pictures.

Theoretical sampling is a special type of purposeful sampling guided by the attempt of the researcher “to seek statements, events, or cases” (Charmaz, 2006, p. 103) which can assist the researcher in interpreting the already emerged concepts or categories. Theoretical sampling in my research had strong connections to constant comparison and to reliance on existing theories and bodies of knowledge. During theoretical sampling, I compared the newly collected data with the already analysed and categorised items. In the GTM literature, these repeated steps are called constant comparison, and the process is defined as “the act of comparing one piece of data (…) with another piece of data (…) attached to the same concept in order to see if it represents the same concept” (Urquhart, 2013, p. 192). In addition to the constant comparison of data, I repeatedly turned to the disciplinary literature to find out whether cases and events similar to the ones I was examining had already been documented by previous research. I did not limit my reading to the literature in the linguistics and SLA disciplinary area. I adopted a noncommittal position determined to keep openness

and leap disciplinary boundaries. Consequently, I considered scientific evidence and theories from a broad range of research areas to keep my data analysis theoretically informed.

The action of constant comparison is fairly complex, labour intensive, and it permeates the whole research process situated within the GTM paradigm. It can be likened to a sophisticated mechanism that processes and groups data. To illustrate how I performed the action of constant comparison of data in my study, I use the data-engine metaphor (Dick, 2007, pp. 408–409) shown in Figure 5. I modified Dick’s (2007) original idea to suit the purposes of my study but kept his concepts of “agreement”, “disagreement”, and

“explanation” (Dick, 2007, p. 408) along with their original content. In what follows, I present the steps of the constant comparison process as they took place with the help of the data engine, then I give some examples of how data was compared and how, on certain occasions, it underwent some transformations.

In the process of constant comparison of data, I followed the steps listed below:

a. I compared items of data within the sets of data. For example, I contrasted lines from the pre-task interview collected from different participants. Then, I compared different sets of data to each other. For example, I compared the set of data originating from the pre-task interviews to the sets of data collected in the think-aloud activity.

Subsequently, I compared the set of pre-task interview data to the set of post-task interview data and to the set of data gained from the think-aloud activity. At a more advanced stage of data analysis, the comparison involved contrasting categories with the sets of data collected in the different phases of the research.

b. The aim of data comparison throughout the analysis was the grouping of the data based on agreement and disagreement patterns. An agreement pattern meant the existence of identical or very similar content in the data. I used the coding label consistence across data from different sources to mark this type of data. A disagreement pattern meant the existence of divergent and dissimilar content in the data. I used the coding label divergence across data to mark these instances.

c. In the case of agreement or consistence across fairly large amounts of data, I took further steps to identify exceptions. I considered exceptions to be disagreement patterns. Disagreement patterns are important constituents? in the process of category construction as they have the potential to help the researcher develop a more nuanced understanting of the issue under scrutiny.

d. In the case of disagreements, I took further action to find possible explanations for their existence. This action involved turning to the participants (with the occasion of the post-task interview) and searching specialist literature.

Figure 5. The Constant Comparison Data-Engine.

Collected data was compared intra- and inter-set. Comparison aimed at identifying agreeing and diverging data patterns. Agreeing data patterns were coded agreement. I re-examined agreeing data patterns in comparison to other data or to existent literature. Data element coded agreement became constituents of emergent theory. I also compared the emergent theory to exiting sets of data or newly collected data. Constant comparison aimed also at identifying exceptions. Identified exceptions were coded disagreement. To discover why they occurred, I undertook further investigation. This took the form of data collection from the participants and/or the consultation of disciplinary literature. These actions resulted in new data sets and/or emergent theory. The whole process was iterative and grounded both in data (empirically) and literature (theoretically). Adapted from “What can grounded theorists and action researchers learn from each other”, by B. Dick 2007. p. 408. Los Angeles, CA. SAGE

e. The post-task interview sessions offered favourable opportunities to share my data interpretation with the participants. Their feedback acted as a form of new “’slices’ of data” (Glaser & Strauss, 1967, p. 61; Urquhart, 2019, p. 98) and contributed to the development of the properties of the categories.

During the steps of theoretical sampling and constant comparison, I included several of the participants’ statements in more than one category. When pieces of data are interpreted in the light of more than one theoretical possibilities (Glaser & Strauss, 1967, p. 105), the operation of coding is called tentative. For example, the concept of frequency of reliance on a language came into being at the end of a coding process that started with the operation of tentative coding. Figure 6 illustrates how the concept frequency of reliance of use, at a point in my constant comparison analysis, became a property ascribed to the subcategory

“shutting down” languages because I interpreted it as a result of the language user’s attitude towards the languages he knew. However, at a later point in the research, I changed its place, and it became one of the properties of the subcategory using languages “silently”. I opted for this rearrangement because it became obvious that the participants explained the term ‘frequency’ as having a positive connotation and linked it to the number of occasions they made use of a certain language and not to the occasions when they had to keep their linguistic systems separate.

Figure 6. Constant Comparison of Data Often Results in the Rearrangment of Properties, Subcategories and Categories.

Hexagons denote initial concepts. Bold-contoured rectangles denote subcategories. The oval denotes the category. The curved arrow represents the movement of the concept frequency of use. Words between quotation marks are in vivo codes.

The operations of rethinking and rearranging the interpretations of the collected data played a central role in the actions of theoretical sampling and constant comparison and were instrumental for the operation of coding. Charmaz (2006) and Wiener (2007) describe coding as the close and careful examination of data with the aim of discovering the nature of certain processes, states or events, and the relationships between them. Codes are empirical observations transposed into theoretical terms (Bryant & Charmaz, 2007). The main characteristics of codes are their impermanence and abstractness. In the process of data analysis, codes become modified as they are taken to higher level of abstractness or as new attributes are attached to them. I coded the research data in several sessions. During the early phases of data analysis, I coded by hand, using highlighter pens, paper note stickers and tables in which I organized data. Only later did I start to use the Quirkos qualitative data analysis software and reanalysed the collected data for a third time. Figure 7 is an example of coding configuration obtained with this software.

Figure 7. Coding Configuration Chart Based on the Answer of a Research Participant About Learning and Using Multiple Languages.

Coloured dots represent categories and subcategories. Subcategories are smaller in size and are arranged around categories. The colours in the body of the transcribed text correspond to the colours of the categories and subcategories.

The mechanics of coding I used in the present research are described later. Here I explain how the coding operation benefited from using in vivo codes and how the operation of writing memos assisted the process of data analysis, and then I give some examples of data from interviews in order to allow the reader a better insight into the process of data analysis.

In vivo codes are terminology used by the participants, the term “in vivo” meaning

“occurring in the natural setting”. As already mentioned, codes are abstractions rooted in empirical observations. They can materialize either as substantive definitions formulated by the researcher or as in vivo terms encapsulating the emic perspective. The adoption of in vivo codes in the analysis of data maintains the insiders’ voice, often condensing meaning in metaphorical and ingenious ways and relating the analysis to ordinary life, keeping the research study grounded in data (Charmaz, 2006). Interview data yielded several metaphors and analogies that could be abstracted to the level of in vivo codes. The interviews were conducted in Hungarian. In the examples below, the original terms are quoted in Hungarian and then translated into English. The participants’ names are reduced to initials.

Participants often described the languages they knew but did not use in certain situations as being “a színfalak mögött”, that is “backstage” (participant B.Cs). They also explained the “backstage” metaphor as a state of “másoknak láthatatlan”—“invisible to others”

(participant B. Cs), and “észrevehetetlen”—“which goes unnoticed” (participant H. B). In the analysis, the “backstage” metaphor became a code with two dimensions: languages that are not relied on at all and languages used as main suppliers and helpers in the performance of particular linguistic tasks. For both these dimensions, which later evolved into categories, I used the in vivo codes borrowed from the participants’ discourse.

The action of constant comparison was backed up substantially by theoretical memo-writing. In GTM, memos are primarily notes which support substantially the discerning process of the analysis (Birks & Mills, 2012; Charmaz, 2006; Urquhart, 2013). They most usually contain the observations, thoughts, remarks of the researcher connected to the various issues of the research process. In this, they are close to mundane personal notes written as reminders. Theoretical memo writing appears in the GTM also as an essential operation which assists the researcher in being self-reflective, starting to theorize and in “being practical with theory” (Hasan, 2014, title), building the steps and the strategies to be used in the research process. Some GTM theoreticians consider memo writing an indispensable operation in the GTM research paradigm and advise researchers to write as much as they can (Birks & Mills, 2012; Charmaz, 2006; Stern, 2007; Urquhart, 2013). In conducting my research, theoretical memo writing became a habit as soon as I discovered that the process of jotting down my ideas related to the emerging concepts, the relationships between them, their theoretical underpinnings was helpful in both thinking forward and in keeping track of what I had already done. Theoretical memo writing helped me see more clearly the greater picture of concepts

Figure 8. Edited Theoretical Memo on Defining the Construct of Proficiency.

DEBRIEFING PROFICIENCY & KNOWING WORDS & ASSESSING PROFICIENCY How does knowing words relate to PROFICIENCY?

WORKING DEFINITION/TENTATIVE DEFINITION OF PROFICIENCY

There are still some quiet grumblings about how to define and what to include when defining proficiency in a language. Assessments of language proficiency are nowadays very popular (especially) in Hungary. However, their usefulness has not been demonstrated in the sense that there is no convincing literature on whether these assessments objectively evaluate test takers’ language mastery. Then, it is very often the case that the language user considers the results of the proficiency assessment as not reflecting (their) true abilities. Plus, assessments of language proficiency test the academic vocabulary required by academic discourse. Plus, testing professionals and language educators still embrace the impractical view of THE PERFECT MONOLINGUAL Æ the well-educated one-language-user, who uses a high-standard written and spoken language. This view has spread from educators/testing professional to language users as well, and it is common practice to consider and contrast own language knowledge with this popularized highly able, flawless intellectual speaker. It was a surprise for me to discover that some participants tend to subject their native language to this scrutiny.

Transylvanian-Hungarian L1 participants living in the NW part of Romania, in an L2 Romanian environment, tend to view their native language vocabulary as:

“nem megfelelő”—“inadequate” (N.A.)

“sok kivánnivalót hagy maga után”—“leaves much to be desired” (F.L)

“hiányos”—“deficient” (S.S)

“alkalmatlan”—“inapt” (P. F) (interviews, 2014)

This view of proficiency disregards the changing nature of proficiency and competence. It disregards the fact that they are always in a state of activity. Proficiency and competence are both filled with activity. They keep moving and undergoing change. In addition, measures of proficiency and competence capture a static picture of a language user’s performance limited to a certain point in time, certain environment and certain types of linguistic tasks. Martin Seligman (former president of the American Psychology Association & co-founder of the positive psychology movement) remarked that “language tests were developed to separate those test takers who cannot achieve the requirements of the CEF from those who are able to appropriate the standardized requirements. The tests are not designed to find the strengths within either group” (Sommer & Sommer, 2002, p. 225).

The moral to be drawn from the above discussion on the construct of proficiency and its relation to the individual’s lexis is that proficiency cannot be condensed into a single definition because it rests on multiple aspects. It is a knotty concept, a so-called “big fuzzy” (Sommer & Sommer, 2002, p. 218).

Literature read:

Sommer, R., & Sommer, B. (2002). A practical guide to behavioral research. Tools and techniques. 5^th edition. Oxford: OUP.

Malinowski, B. (1923). The problem of meaning in primitive languages. In C. K. Ogden &

I. A. Richards (Eds.), The meaning of meaning. New York: Hartcourt Brace & World. pp.

296-335.

Cui, X. (2012). ‘How are you?’--‘Fine, thanks. How about you?’: A case of problematic social interaction at work between Chinese and Australians. In C. Gitsaki & R. B. Baldauf (Eds.), Future directions in applied linguistics: Local and global perspectives (pp. 373-389; Chapter 21). Cambridge: CUP.

and categories and the ways they became integrated in relational structures. My memos have taken the form of research logs and several memo texts have become incorporated in the text of the present thesis.

The memo in Figure 8 reproduces an edited reflective commentary written while I was trying to define the construct of proficiency related to lexical knowledge in a way which would prove flexible enough to accommodate the transitory characteristics of lexical retrieval ability.

That is, the inclusion of the notion that assessed lexical knowledge and an individual’s ability to retrieve words from the mental lexicon is by default unbalanced across contexts and time.

While I was interviewing the L1 Transylvanian-Hungarian participants, I was surprised to discover that many of them were critical of their abilities in L1 Hungarian, and some characterized their lexical knowledge as being inadequate and insufficient for academic purposes. This prompted me to both revisit the literature on language proficiency issues and ask the participants to give their own definitions of what knowing a word meant for them. This investigation triggered further theoretical memo writing, which later led to the refining of some categories which were included in the theoretical framework of the study, and which in turn supported the process of theorizing.