Volume 5, No. 2, Art. 4 – May 2004
Remodeling Grounded Theory
Barney G. Glaser with the assistance of Judith Holton
Abstract: This paper outlines my concerns with Qualitative Data Analysis' (QDA) numerous remodelings of Grounded Theory (GT) and the subsequent eroding impact. I cite several examples of the erosion and summarize essential elements of classic GT methodology. It is hoped that the article will clarify my concerns with the continuing enthusiasm but misunderstood embrace of GT by QDA methodologists and serve as a preliminary guide to novice researchers who wish to explore the fundamental principles of GT.
Key words: grounded theory, qualitative data analysis, constant comparative method, theoretical sensitivity
Table of Contents
2. QDA Blocking of GT
3. Grounded Theory Procedures
3.1 Theoretical sensitivity
3.2 Getting started
3.3 All is data
3.4 Use of the literature
3.5 Theoretical coding
3.6 Open coding
3.7 Theoretical sampling
3.8 Constant comparative method
3.9 Core variable
3.10 Selective coding
3.12 Interchangeability of indicators
3.15 Sorting and writing up
3.16 Analytic rules developed during sorting
The difference between the particularistic, routine, normative data we all garner in our everyday lives and scientific data is that the latter is produced by a methodology. This is what makes it scientific. This may sound trite, but it is just the beginning of many complex issues. Whatever methodology may be chosen to make an ensuing research scientific has many implicit and explicit problems. It implies a certain type of data collection, the pacing and timing for data collection, a type of analysis and a specific type of research product. 
In the case of qualitative data, the explicit goal is description. The clear issue articulated in much of the literature regarding qualitative data analysis (QDA) methodology is the accuracy, truth, trustworthiness or objectivity of the data. This worrisome accuracy of the data focuses on its subjectivity, its interpretative nature, its plausibility, the data voice and its constructivism. Achieving accuracy is always worrisome with a QDA methodology. 
These are a few of the problems of description. Other QDA problems include pacing of data collection, the volume of data, the procedure and rigor of data analysis, generalizability of the unit findings, the framing of the ensuing analysis and the product. These issues and others are debated at length in the qualitative research literature. Worrisome accuracy of qualitative data description continually concerns qualitative researchers and their audiences. I have addressed these problems at length in "The Grounded Theory Perspective: Conceptualization Contrasted with Description" (GLASER, 2001). 
In this paper I will take up the conceptual perspective of classic Grounded Theory (GT). (In some of the research literature, classic GT methodology has also been termed GLASERian GT although I personally prefer the term "classic" as recognition of the methodology's origins.) The conceptual nature of classic GT renders it abstract of time, place and people. While grounded in data, the conceptual hypotheses of GT do not entail the problems of accuracy that plague QDA methods. 
The mixing of QDA and GT methodologies has the effect of downgrading and eroding the GT goal of conceptual theory. The result is a default remodeling of classic GT into just another QDA method with all its descriptive baggage. Given the ascending focus on QDA by sheer dint of the number of researchers engaged in qualitative analysis labeled as GT, the apparent merger between the two methodologies results in default remodeling to QDA canons and techniques. Conceptual requirements of GT methodology are easily lost in QDA problems of accuracy, type data, constructivism, participant voice, data collection rigor according to positivistic representative requirements, however couched in a flexibility of approach (see LOWE, 1997). The result is a blocking of classic GT methodology and the loss of its power to transcend the strictures of worrisome accuracy—the prime concern of QDA methods to produce conceptual theory that explains fundamental social patterns within the substantive focus of inquiry. 
I will address some, but not all, of the myriad of remodeling blocks to classic GT analysis brought on by lacing it with QDA descriptive methodological requirements. My goal is to alleviate the bane on good GT analysis brought on by those QDA senior researchers open to no other method, especially the GT method. I hope to relieve GT of the excessive scientism brought on it by those worried about accuracy and what is "real" data when creating a scientific product. I hope to give explanatory strength to those PhD dissertation level students to stand their GT grounds when struggling in the face of the misapplied QDA critique by their seniors and supervisors. 
I wish to remind people, yet again, that classic GT is simply a set of integrated conceptual hypotheses systematically generated to produce an inductive theory about a substantive area. Classic GT is a highly structured but eminently flexible methodology. Its data collection and analysis procedures are explicit and the pacing of these procedures is, at once, simultaneous, sequential, subsequent, scheduled and serendipitous, forming an integrated methodological "whole" that enables the emergence of conceptual theory as distinct from the thematic analysis characteristic of QDA research. I have detailed these matters in my books "Theoretical Sensitivity" (GLASER, 1978), "Basics of Grounded Theory Analysis" (GLASER, 1992), "Doing Grounded Theory" (GLASER, 1998a), and "The Grounded Theory Perspective" (GLASER, 2001). 
Over the years since the initial publication of "Discovery of Grounded Theory" (GLASER & STRAUSS, 1967), the transcendent nature of GT as a general research methodology has been subsumed by the fervent adoption of GT terminology and selective application of discrete aspects of GT methodology into the realm of QDA research methodology. This multi-method cherry picking approach, while obviously acceptable to QDA, is not compatible with the requirements of GT methodology. 
Currently it appears to be very popular in QDA research substantive and methodological papers to label QDA as GT for the rhetorical legitimating effect and then to critique its various strategies as somewhat less than possible or effective; then further, to sanctify the mix of methods as one method. Classic GT is not what these "adopted QDA" usages would call GT. These researchers do not realize that while often using the same type of qualitative data, the GT and QDA methods are sufficiently at odds with each other as to be incapable of integration. Each method stands alone as quite legitimate. The reader is to keep in mind that this paper is about GT and how to extract it from this remodeling. It does not condemn QDA in any way. QDA methods are quite worthy, respectable and acceptable. As I have said above, the choice of methodology to render research representations about qualitative data as scientific is the researcher's choice. But there is a difference between received concepts, problems and frameworks imposed on data by QDA methods and GT's focus on the generation and emergence of concepts, problems and theoretical codes. The choice of methodology should not be confused, lumped or used piecemeal if GT is involved. To do so is to erode the conceptual power of GT. 
As such, GT procedures and ideas are used to legitimate and buttress routine QDA methodology. Considering the inundation, overwhelming and overload of QDA dictums, "words" and assumed requirements on GT methodology, the reader will see that it is hard to both assimilate and withstand this avalanche on GT methodology. The assault is so strong and well meaning that many—particularly novice researchers—do not know, nor realize, that GT is being remodeled by default. 
The view of this paper is that the researcher who has to achieve a GT product to move on with his or her career and skill development is often blocked by the confusion created through this inappropriate mixing of methods and the attendant QDA requirements thus imposed. Undoing the blocks to GT by this default remodeling will not be an easy task given the overwhelming confusion that has resulted and seems destined to continue to grow. 
I will deal with as many of the blocks as I see relevant but certainly not all. If I repeat, it will be from different vantage points to undo QDA remodeling in the service of advancing the GT perspective. I will hit hard that GT deals with the data as it is, not what QDA wishes it to be or, more formally, what QDA preconceives to be accurate and to be forcefully conceptualized. This requires honesty about taking all data as it comes, figuring it out and then its conceptualization. I have written at length on "all is data" and on forcing in "Doing Grounded Theory" (GLASER, 1998a). 
As I deal with this escalating remodeling of GT to QDA requirements, my hope is to free GT up to be as originally envisioned. In "Theoretical Sensitivity" I wrote: "The goal of grounded theory is to generate a conceptual theory that accounts for a pattern of behavior which is relevant and problematic for those involved. The goal is not voluminous description, nor clever verification." (GLASER, 1978, p.93) 
This paper has a simple message. GT is a straightforward methodology. It is a comprehensive, integrated and highly structured, yet eminently flexible process that takes a researcher from the first day in the field to a finished written theory. Following the full suite of GT procedures based on the constant comparative method, results in a smooth uninterrupted emergent analysis and the generation of a substantive or formal theory. When GT procedures are laced with the exhaustive, abundant requirements of QDA methodology, GT becomes distorted, wasting large amounts of precious research time and derailing the knowledge—hence grounding—of GT as to what is really going on. The intertwining of GT with preconceived conjecture, preconceptions, forced concepts and organization, logical connections and before-the-fact professional interest defaults GT to a remodeling of GT methodology to the status of a mixed methods QDA methodology. This leads to multiple blocks on conceptual GT. 
The word "analysis" is a catchall word for what to do with data. It is "scientized" up, down and sideways in QDA methodologies catching up GT analysis in its wake. QDA leads to particularistic analysis based on discrete experiences while blocking the abstract idea of conceptualizing latent patterns upon which GT is based. When GT becomes laced with QDA requirements, it is hard to follow to the point of confusion. Theory development is confused with QDA description thereby blocking GT generation of conceptual theory. 
GT has clear, extensive procedures. When brought into QDA, GT abstraction is neglected in favor of accuracy of description—the dominant concern of QDA methodology—and GT acquires the QDA problem of worrisome accuracy—an irrelevant concern in GT. To repeat, GT methodology is a straightforward approach to theory generation. To spend time worrying about its place in QDA methods and science is just fancy, legitimating talk, but the result is the defaulting of GT to the confusion of QDA analysis. 
CRESWELL in his book "Qualitative Inquiry and Research Design" (1998) lumps GT into comparisons with phenomenology, ethnography, case study and biographical life history. The result of the lumping is a cursory default remodeling of GT to a "kind" of QDA. This lumping of GT with other QDA methods prevents GT from standing alone as a transcending general research methodology. The criteria of CRESWELL's continuum organize methods according to when theory is used in research, varying from before the study begins to post-study. By study, he means data collection and structuring questions. This is a very weak gradation for discerning the difference among QDA methods and GT methodology. CRESWELL clearly does not discern the difference between generating theory from data collection and generating theory that applies to the data once collected. Both come during and after data collection, but are very differently sourced. The result is a lumping and confusion of GT with QDA. 
CRESWELL (1998, p.86) says:
"At the most extreme end of the continuum, toward the 'after' end, I place grounded theory. Strauss and Corbin (1990) are clear that one collects and analyzes data before using theory in a grounded theory study. This explains, for example, the women's sexually abuse study by Morrow and Smith (1995) in which they generate the theory through data collection, pose it at the end, and eschew prescribing a theory at the beginning of the study. In my own studies, I have refrained from advancing a theory at the beginning of my grounded theory research, generated the theory through data collection and analysis, posed the theory as a logic diagram and introduced contending and contrasting theory with the model I generate at the end of my study (Creswell & Brown 1992, Creswell and Urbom 1997)." 
CRESWELL may be stating a fundamental tenant of GT—begin with no preconceived theory and then generate one during the analysis (unless he meant applying an extant theory). As a distinguishing item of GT, however, it is barely a beginning, leaving the reader with no knowledge of how generating is done, because the assumption is that it is done by routine QDA. Contrasting the generated theory with extant other theories to prove, improve or disprove one or the other neglects or ignores constantly comparing the theories for category and property generation. This contrasting with other theories also prevents modifying the GT generated theory using the other theory as a kind of data. Both constant comparing and modifying are two vital tenants of GT. 
GT may or may not be mentioned in a QDA methodological discussion, but its procedures frequently are. As such, constant comparative analysis, problem emergence, theoretical sampling, theoretical saturation, conceptual emergence, memoing, sorting, etc. become laced with QDA requirements thereby defaulting their rigorous use to a QDA burden. This virtual subversion of GT results in complex confusion of an otherwise simple methodology for novice researchers. The researcher is blocked and no longer freed by the power and autonomy offered by GT to arrive at new emergent, generated theory. The ability to be honest about what exactly is the data is consequently distorted by the unattainable quest for QDA accuracy. For example, Kathryn MAY unwittingly erodes the GT methodology in QDA fashion when describing the cognitive processes inherent in data analysis.
"Doing qualitative research is not a passive endeavor. Despite current perceptions and student's prayers, theory does not magically emerge from data. Nor is it true that, if only one is patient enough, insight wondrously enlightens the researcher. Rather, data analysis is a process that requires astute questioning, a relentless search for answers, active observation, and accurate recall. It is a process of piecing together data, of making the invisible obvious, of recognizing the significant from the insignificant, of linking seemingly unrelated facts logically, of fitting categories one with another, and of attributing consequences to antecedents. It is a process of conjecture and verification, of correction and modification, of suggestion and defense. It is a creative process of organizing data so that the analytic scheme will appear obvious." (MAY, 1994, p.10) 
Dr MAY engages in descriptive capture in QDA fashion and attacks the main tenant of GT, that theory can emerge. She is lost in accurate fact research, which is moot for GT. She prefers to force the data, making it obey her framework. She does not acknowledge the constant comparative method by which theory emerges from all data. Again, GT is defaulted to routine QDA. 
Similarly, this PhD student—in her e-mail cry to me for help—wanted to do a GT dissertation but was caught up in QDA and descriptive capture.
"I need some guidance. I'm on wrong track—I don't care about the main concerns of clinical social workers in private practice. I care about the main concerns of anyone attempting to contextualize practice. Maybe the issue is that I'm interested in an activity regardless of the actor. If I ask these questions I have no doubt that main concerns will emerge as well as attempts to continually resolve them. This I care about." (e-mail correspondence, Jan 2002) 
She is caught by the QDA approach to force the data for a professional concern. She wants to use GT procedures in service of a QDA forcing approach, which defaults GT. GT does not work that way, but the prevalence of QDA would have her think that way. Later, under my guidance, she let the main concern emerge and did an amazingly good dissertation on binary deconstruction between social worker and client. 
The GT problem and core variable must emerge and it will. I have seen it hundreds of times. Later, when the GT's main concern emerges and is explained in a generated theory, it will have relevance for professional concerns. Starting before emergence with the professional interest, a problem is very likely to result in research with little or no relevance in GT—just routine QDA description with "as if" importance. 
Here is a good example of extensive lacing of GT by QDA needs. The confusion of QDA requirements and GT procedures, in this example, makes it hard to follow and clearly erodes GT by default remodeling.
"Comprehension is achieved in grounded theory by using tape-recorded, unstructured interviews and by observing participants in their daily lives. However, the assumption of symbolic interactionism that underlie grounded theory set the stage for examining process, for identifying stages and phases in the participant's experience. Symbolic interaction purports that meaning is socially constructed, negotiated and changes over time. Therefore the interview process seeks to elicit a participant's story, and this story is told sequentially as the events being reported unfold. Comprehension is reached when the researcher has interviewed enough to gain in-depth understanding." (MORSE, 1994, p.39) 
In fact, GT does not require tape-recorded data. Field notes are preferable. GT uses all types of interviews and, as the study proceeds, the best interview style emerges. It is not underlined by symbolic interaction, nor constructed data. GT uses "all as data," of which these are just one kind of data. GT does not preconceive the theoretical code of process. There are over 18 theoretical coding families of which process is only one. In GT, its relevance must emerge; it is not presumed. Interviews lead to many theoretical codes. Participant stories are moot. Patterns are sought and conceptualized. GT does not search for description of particularistic accounts. All data are constantly compared to generate concepts. 
MORSE continues her description of GT:
"Synthesis is facilitated by adequacy of the data and the processes of analysis. During this phase the researcher is able to create a generalized story and to determine points of departure, of variation in this story. The process of analysis begins with line-by-line analysis to identify first level codes. Second-level codes are used to identify significant portions of the text and compile these excerpts into categories. Writing memos is key to recording insight and facilitates, at an early stage, the development of theory." (MORSE, 1994, p.39 
It is, indeed, hard to recognize GT procedures in this quote by MORSE. "Adequacy of data" and a "generalized story" smack of worrisome accuracy and descriptive capture, which are pure QDA concerns. They do not relate to GT procedures. GT fractures the story in the service of conceptualization. Her approach to line-by-line analysis is a bare reference to the constant comparative process, but that is all. Her references to first level, second level codes, portions of text and compiling excerpts into categories are far from the constant comparative method designed to generate conceptual categories and their properties from the outset of data collection and analysis. Writing memos in GT has to do with immediate recording of generated theoretical conceptual ideas grounded in data, not the mystical—perhaps conjectural—insights to which MORSE refers to. 
MORSE continues with her description of GT:
"As synthesis is gained and the variation in the data becomes evident, grounded theorists sample according to the theoretical needs of the study. If a negative case is identified, the researcher, theoretically, must sample for more negative cases until saturation is reached when synthesis is attained." (MORSE, 1994, p.39) 
Again, finding GT procedures in this description is hard. There is always variation in the data. GT is concerned with generating a multivariate conceptual theory—not data variation for QDA. In GT, seeking negative cases is not a procedure. This is more likely to be preconceived forcing. GT seeks comparative incidents by theoretical sampling. The purpose in sampling is to generate categories and their properties. The GT researcher does not know in advance what will be found. Incidents sampled may be similar or different, positive or negative. MORSE's reference to saturation does not imply conceptual saturation; rather, it anticipates simple redundancy without conceptual analysis. 
"Theorizing follows from the processes of theoretical sampling. Typologies are constructed by determining two significant characteristics and sorting participants against each characteristic on a 2x2 matrix. Diagramming is used to enhance understanding and identifying the basic social process (BSP) that accounts for most of the variation in the data." (MORSE, 1994, p.39). 
Theorizing in GT is an emergent process generated by continuous cycling of the integrated processes of collecting, coding and conceptual analysis with the results written up constantly in memos. Theoretical sampling is just one source of grounding during the constant comparative method. Preconceiving theoretical codes such as typologies or basic social processes (BSPs) is not GT. In GT, relevant theoretical codes emerge in conceptual memo sorting and could be "whatever." While the fourfold property space is a good tool, when emergent, for conceptualizing types (see GLASER & STRAUSS, "Awareness of Dying," 1965), it is not for placing or sorting participants, a priori, nor for counting them. This is strictly routine, preconceived QDA descriptive capture, not GT. 
"As with the methods previously discussed, recontextualization is determined by the level of abstraction attained in the model development. Whereas substantive theory is context bound, formal theory is more abstract and may be applicable to many settings or other experiences." (MORSE, 1994, p.34) 
This statement is totally wrong for GT, but it addresses the usual QDA quandary of trying to generalize a description of a unit. In contrast, GT substantive theory always has general implications and can easily be applied to other substantive areas by the constant comparative method of modifying theory. For example, by comparing incidents and modifying the substantive theory of milkmen who engage in cultivating housewives for profit and recreation, a GT of cultivation can apply easily to doctors cultivating clients to build a practice, thereby expanding the original substantive theory to include cultivating down instead of cultivating up the social scale. Formal theory is generated by many such diverse area comparisons done in a concerted way to generate a formal theory of cultivating for recreation, profit, client building, help, donations etc. 
Context must emerge as a relevant category or as a theoretical code like all other categories in a GT. It cannot be assumed as relevant in advance. As one applies substantive theory elsewhere or generates formal theory, context—when relevant—will emerge. 
These quotes clearly lump GT into the multi-method QDA camp with the result being default remodeling by erosion of classic GT methodology. Nowhere does MORSE refer to the GT procedures of delimiting at each phase of generating, of theoretical completeness, conceptual saturation, core variable analysis, open to selective coding, memo banks, analytic rules, theoretical sorting, memo piles writing up, reworking and resorting, emergent problem, interchangeability of indices and theoretical (not substantive) coding. The effect of such default remodeling is a great loss of essential GT procedures blocked by the imposition of QDA worrisome accuracy requirements. 
GT requires following its rigorous procedures to generate a theory that fits, works, is relevant and readily modifiable. When it is adopted, co-opted, and corrupted by QDA research, a close look at the work often shows that the QDA researcher is tinkering with the GT method. He or she brings it into a QDA research design to comply with the strictures and professional expectations of the dominant paradigm. Getting some kind of product with a few concepts rescues the QDA research, since the QDA description alone does not suffice. Then, the GT label is used to legitimate the QDA research. 
GT stands alone as a conceptual theory generating methodology. It is a general methodology. It can use any data, but obviously the favorite data, to date, is qualitative data. Ergo GT is drawn into the QDA multi-method world and eroded by consequence, however unwittingly. This revealing of method muddling (see BAKER, WUEST, & STERN, 1992) of procedures does a tinkering rescue job, but the result is that GT is default remodeled. GT becomes considered, wrongly, as an interpretative method, a symbolic interaction method, a constructionist method, a qualitative method, a describing method, a producer of worrisome facts, a memoing method, an interview or field method and so forth. It is clear that this tinkering by QDA researchers indicates they are too derailed by QDA to learn systematic GT procedures. At best, a few GT procedures are borrowed out of context. 
These above authors are typical of many trying to place GT somewhere in the QDA camp. First they lace it with some QDA requirements and ideas, which they then use to lump GT into QDA multi-method thought. Lumping GT in as a QDA methodology simply does not apply and, indeed, blocks good GT while the default remodeling of GT into another QDA rages on. Lumping erodes GT. In the remainder of this article, I will try to show how GT stands alone on its own, as a conceptualizing methodology. My goal will be to bring out the classic GT perspective on how GT analysis is done—to lay this method bare—and in the bargain to show how QDA blocks, as I have said, GT generation and product proof. 
When not laced and lumped with QDA requirements, GT procedures are fairly simple. The blocking problems come with the method mixing. I have already written in detail much about GT procedures in "Discovery of Grounded Theory" (GLASER & STRAUSS, 1967), "Theoretical Sensitivity" (GLASER, 1978), "Doing Grounded Theory" (GLASER, 1998a), "Basics of Grounded Theory Analysis (GLASER, 1992), "More Grounded Theory Methodology" (GLASER, 1994), and "The Grounded Theory Perspective" (GLASER, 2001), all by Sociology Press. I have also published many examples of a "good" GT analysis—"Examples of Grounded Theory" (GLASER, 1993), "Grounded Theory 1984 to 1994" (GLASER, 1995), "Gerund Grounded Theory" (GLASER, 1998b)—and have given many references in my books. 
The GT product is simple. It is not a factual description. It is a set of carefully grounded concepts organized around a core category and integrated into hypotheses. The generated theory explains the preponderance of behavior in a substantive area with the prime mover of this behavior surfacing as the main concern of the primary participants. I have said over and over that GT is not findings, not accurate facts and not description. It is just straightforward conceptualization integrated into theory—a set of plausible, grounded hypotheses. It is just that—no more—and it is readily modifiable as new data come from whatever source—literature, new data, collegial comments, etc. The constant comparative method weaves the new data into the sub-conceptualization. What is important is to use the complete package of GT procedures as an integrated methodological whole. 
The following is a summary of the essential elements of GT methodology: Bear in mind, when reading this summary, that the goal of GT is conceptual theory abstract of time, place and people. The goal of GT is NOT the QDA quest for accurate description. 
The ability to generate concepts from data and to relate them according to normal models of theory in general, and theory development in sociology in particular, is the essence of theoretical sensitivity. Generating a theory from data means that most hypotheses and concepts not only come from the data, but are systematically worked out in relation to the data during the course of the research. A researcher requires two essential characteristics for the development of theoretical sensitivity. First, he or she must have the personal and temperamental bent to maintain analytic distance, tolerate confusion and regression while remaining open, trusting to preconscious processing and to conceptual emergence. Second, he/she must have the ability to develop theoretical insight into the area of research combined with the ability to make something of these insights. He/she must have the ability to conceptualize and organize, make abstract connections, visualize and think multivariately. The first step in gaining theoretical sensitivity is to enter the research setting with as few predetermined ideas as possible—especially logically deducted, a prior hypotheses. The research problem and its delimitation are discovered. The pre-framework efforts of QDA block this theoretical sensitivity. 
A good GT analysis starts right off with regular daily data collecting, coding and analysis. The start is not blocked by a preconceived problem, a methods chapter or a literature review. The focus and flow is immediately into conceptualization using the constant comparative method. The best way to do GT is to just do it. It cannot fail as the social psychological world of structure, culture, social interaction, social organization etc. goes on irrespective. There always is a main concern and there always is a prime mover. As an open, generative and emergent methodology, GT provides an honest approach to the data that lets the natural organization of substantive life emerge. The GT researcher listens to participants venting issues rather than encouraging them to talk about a subject of little interest. The mandate is to remain open to what is actually happening and not to start filtering data through pre-conceived hypotheses and biases to listen and observe and thereby discover the main concern of the participants in the field and how they resolve this concern. The forcing, preconceived notions of an initial professional problem, or an extant theory and framework are suspended in the service of seeing what will emerge conceptually by constant comparative analysis. When QDA requires this preconception, GT is rendered non-emergent through coding and memoing as the researcher tries to follow a non-emergent problem. 
GT stands alone as a conceptual theory generating methodology. It can use any data, but obviously the favorite data to date is qualitative. While interviews are the most popular, GT works with any data—"all is data"—not just one specific data. It is up to the GT researcher to figure out what data they are getting. The data may be baseline, vague, interpreted or proper-line. The data is not to be discounted as "subjective," "obvious," "constructed," etc, as we find in QDA critiques. There is always a perception of a perception as the conceptual level rises. We are all stuck with a "human" view of what is going on and hazy concepts and descriptions about it. GT procedures sharpen the generated concepts systematically. 
It is critical in GT methodology to avoid unduly influencing the pre-conceptualization of the research through extensive reading in the substantive area and the forcing of extant theoretical overlays on the collection and analysis of data. To undertake an extensive review of literature before the emergence of a core category violates the basic premise of GT—that being, the theory emerges from the data not from extant theory. It also runs the risk of clouding the researcher's ability to remain open to the emergence of a completely new core category that has not figured prominently in the research to date thereby thwarting the theoretical sensitivity. Practically, it may well result in the researcher spending valuable time on an area of literature that proves to be of little significance to the resultant GT. Instead, GT methodology treats the literature as another source of data to be integrated into the constant comparative analysis process once the core category, its properties and related categories have emerged and the basic conceptual development is well underway. The pre-study literature review of QDA is a waste of time and a derailing of relevance for the GT Study. 
The conceptualization of data through coding is the foundation of GT development. Incidents articulated in the data are analyzed and coded, using the constant comparative method, to generate initially substantive, and later theoretical, categories. The essential relationship between data and theory is a conceptual code. The code conceptualizes the underlying pattern of a set of empirical indicators within the data. Coding gets the analyst off the empirical level by fracturing the data, then conceptually grouping it into codes that then become the theory that explains what is happening in the data. A code gives the researcher a condensed, abstract view with scope of the data that includes otherwise seemingly disparate phenomena. Substantive codes conceptualize the empirical substance of the area of research. Theoretical codes conceptualize how the substantive codes may relate to each other as hypotheses to be integrated into the theory. Theoretical codes give integrative scope, broad pictures and a new perspective. They help the analyst maintain the conceptual level in writing about concepts and their interrelations. 
It is in the beginning with open coding—and a minimum of preconception—that the analyst is most tested as to his trust in himself and in the grounded method, his skill to use the method and his ability to generate codes and find relevance. The process begins with line-by-line open coding of the data to identify substantive codes emergent within the data. The analyst begins by coding the data in every way possible—"running the data open." From the start, the analyst asks a set of questions—"What is this data a study of?" "What category does this incident indicate?" "What is actually happening in the data?" "What is the main concern being faced by the participants?" and "What accounts for the continual resolving of this concern?" These questions keep the analyst theoretically sensitive and transcending when analyzing, collecting and coding the data. They force him/her to focus on patterns among incidents that yield codes and to rise conceptually above detailed description of incidents. The analyst codes for as many categories as fit successive, different incidents, while coding into as many categories as possible. New categories emerge and new incidents fit into existing categories. 
Open coding allows the analyst to see the direction in which to take the study by theoretical sampling before he/she has become selective and focused on a particular problem. Thus, when he/she does begin to focus, he/she is sure of relevance. The researcher begins to see the kind of categories that can handle the data theoretically, so that he/she knows how to code all data, ensuring the emergent theory fits and works. Open coding allows the analyst the full range of theoretical sensitivity as it allows him to take chances on trying to generate codes that may fit and work. 
Line by line coding forces the analyst to verify and saturate categories, minimizes missing an important category and ensures the grounding of categories the data beyond impressionism. The result is a rich, dense theory with the feeling that nothing has been left out. It also corrects the forcing of "pet" themes and ideas, unless they have emergent fit. The analyst must do his/her own coding. Coding constantly stimulates ideas. The preplanned coding efforts of routine QDA to suit the preconceived professional problem easily remodel GT by stifling its approach. 
Theoretical sampling is the process of data collection for generating theory whereby the analyst jointly collects, codes and analyses the data and decides what data to collect next and where to find them, in order to develop the theory as it emerges. The process of data collection is controlled by the emerging theory, whether substantive or formal. Beyond the decisions concerning initial collection of data, further collection cannot be planned in advance of the emerging theory. Only as the researcher discovers codes and tries to saturate them by theoretical sampling in comparison groups, do the successive requirements for data collection emerge—both (1) what categories and their properties to be sampled further and (2) where to collect the data. By identifying emerging gaps in the theory, the analyst will be guided as to next sources of data collection and interview style. The basic question in theoretical sampling is to what groups or subgroups does one turn to next in data collection—and for what theoretical purpose? The possibilities of multiple comparisons are infinite and so groups must be chosen according to theoretical criteria. The criteria—of theoretical purpose and relevance—are applied in the ongoing joint collection and analysis of data associated with the generation of theory. As such, they are continually tailored to fit the data and are applied judiciously at the right point and moment in the analysis. In this way, the analyst can continually adjust the control of data collection to ensure the data's relevance to the emerging theory. 
Clearly this approach to data collection done jointly with analysis is far different from the typical QDA preplanned, sequential approach to data collection and management. Imposing the QDA approach on GT would block it from the start. 
The constant comparative method enables the generation of theory through systematic and explicit coding and analytic procedures. The process involves three types of comparison. Incidents are compared to incidents to establish underlying uniformity and its varying conditions. The uniformity and the conditions become generated concepts and hypotheses. Then, concepts are compared to more incidents to generate new theoretical properties of the concept and more hypotheses. The purpose is theoretical elaboration, saturation and verification of concepts, densification of concepts by developing their properties and generation of further concepts. Finally, concepts are compared to concepts. The purpose is to establish the best fit of many choices of concepts to a set of indicators, the conceptual levels between the concepts that refer to the same set of indicators and the integration into hypotheses between the concepts, which becomes the theory. Comparisons in QDA research are between far more general ideas that do not lead to tightly grounded categories. 
As the researcher proceeds to compare incident to incident in the data, then incidents to categories, a core category begins to emerge. This core variable, which appears to account for most of the variation around the concern or problem that is the focus of the study, becomes the focus of further selective data collection and coding efforts. It explains how the main concern is continually resolved. As the analyst develops several workable coded categories, he/she should begin early to saturate as much as possible those that seem to have explanatory power. The core variable can be any kind of theoretical code—a process, a condition, two dimensions, a consequence, a range and so forth. Its primary function is to integrate the theory and render it dense and saturated. It takes time and much coding and analysis to verify a core category through saturation, relevance and workability. The criteria for establishing the core variable within a GT are that it is central, relating to as many other categories and their properties as possible and accounting for a large portion of the variation in a pattern of behavior. The core variable reoccurs frequently in the data and comes to be seen as a stable pattern that is more and more related to other variables. It relates meaningfully and easily with other categories. It has clear and grabbing implications for formal theory. It is completely variable and has conceptual carry through in the emerging theory, enabling the analyst to get through the analyses of the processes that he/she is working on by its relevance and explanatory power. Core variable, conceptual theory is far beyond QDA description or conceptual descriptions which are unending since they are not tied down to a conceptual scheme. A reversion to QDA clearly blocks this necessary theoretical completeness. 
Selective coding means to cease open coding and to delimit coding to only those variables that relate to the core variable in sufficiently significant ways as to produce a parsimonious theory. Selective coding begins only after the analyst is sure that he/she has discovered the core variable. QDA researchers have never figured out the exact purpose and techniques of selective coding. Often they selectively code from the start with preconceived categories. 
Subsequent data collection and coding is thereby delimited to that which is relevant to the emergent conceptual framework. This selective data collection and analysis continues until the researcher has sufficiently elaborated and integrated the core variable, its properties and its theoretical connections to other relevant categories. 
Integrating a theory around a core variable delimits the theory and thereby the research project. This delimiting occurs at two levels—the theory and the categories. First the theory solidifies, in the sense that major modifications become fewer and fewer as the analyst compares the next incidents of a category to its properties. Later modifications are mainly on the order of clarifying the logic, taking out non-relevant properties, integrating elaborating details of properties into the major outline of interrelated categories and—most important—reduction. Reduction occurs when the analyst discovers underlying uniformity in the original set of categories or their properties and then reformulates the theory with a smaller set of higher-level concepts. The second level of delimiting the theory is a reduction in the original list of categories for coding. As the theory grows, becomes reduced, and increasingly works better for ordering a mass of qualitative data, the analyst becomes committed to it. This allows the researcher to pare down the original list of categories for collecting and coding data, according to the present boundaries of the theory. The analyst now focuses on one category as the core variable and only variables related to the core variable will be included in the theory. The list of categories for coding is further delimited through theoretical saturation. Since QDA researchers focus on full description, and no core variable conceptual analysis, delimiting does not occur in QDA research. It just goes on and on—empirical tiny topics draining both researcher and audience. 
GT is based on a concept-indicator model of constant comparisons of incidents (indicators) to incidents (indicators) and, once a conceptual code is generated, of incidents (indicators) to emerging concept. This forces the analyst into confronting similarities, differences and degrees in consistency of meaning between incidents (indicators), generating an underlying uniformity which in turn results in a coded category and the beginnings of properties of it. From the comparisons of further incidents (indicators) to the conceptual codes, the code is sharpened to achieve its best fit while further properties are generated until the code is verified and saturated. 
Conceptual specification, not definition, is the focus of GT. The GT concept-indicator model requires concepts and their dimensions to earn their way into the theory by systematic generation of data. Changing incidents (indicators) and thereby generating new properties of a code can only go so far before the analyst discovers saturation of ideas through interchangeability of indicators. This interchangeability produces, at the same time, the transferability of the theory to other areas by linking to incidents (indicators) in other substantive or sub-substantive areas that produce the same category or properties of it. Interchangeability produces saturation of concepts and their properties, not redundancy of description as some QDA methodologists would have it (see MORSE, 1995, p.147). 
Generating GT takes time. It is above all a delayed action phenomenon. Little increments of coding, analyzing and collecting data cook and mature and then blossom later into theoretical memos. Significant theoretical realizations come with growth and maturity in the data, and much of this is outside the analyst's awareness until preconscious processing becomes conscious. Thus the analyst must pace himself, exercise patience and accept nothing until something happens, as it surely does. Surviving the apparent confusion is important. This requires that the analyst takes whatever amount of quality time that is required to do the discovery process and that he/she learns to take this time in a manner consistent with his/her own temporal nature as an analyst—personal pacing. Rushing or forcing the process will shut down the analyst's creativity and conceptual abilities, exhausting the energy and leaving the researcher empty and the theory thin and incomplete. In QDA work researchers are paced sequentially through the program and framework, and often driven to long periods of no product and exhaustion. To overlay this QDA program on GT severely remodels GT to its deficit. 
Theory articulation is facilitated through an extensive and systematic process of memoing that parallels the data analysis process in GT. Memos are theoretical notes about the data and the conceptual connections between categories. The writing of theoretical memos is the core stage in the process of generating theory. If the analyst skips this stage by going directly to sorting or writing up, after coding, he/she is not doing GT. 
Memo writing is a continual process that leads naturally to abstraction or ideation—continually capturing the "frontier of the analyst's thinking" as he/she goes through data and codes, sorts and writes. It is essential that the analyst interrupts coding to memo ideas as they occur if he/she is to reap the subtle reward of the constant input from reading the data carefully, asking the above questions and coding accordingly. Memos help the analyst to raise the data to a conceptual level and develop the properties of each category that begin to define them operationally. Memos present hypotheses about connections between categories and/or their properties and begin to integrate these connections with clusters of other categories to generate the theory. Memos also begin to locate the emerging theory with other theories with potentially more or less relevance. 
The basic goal of memoing is to develop ideas on categories with complete freedom into a memo fund that is highly sort-able. Memo construction differs from writing detailed description. Although typically based on description, memos raise that description to the theoretical level through the conceptual rendering of the material. Thus, the original description is subsumed by the analysis. Codes conceptualize data. Memos reveal and relate by theoretically coding the properties of substantive codes—drawing and filling out analytic properties of the descriptive data. 
Early on memos arise from constant comparison of indicators to indicators, then indicators to concepts. Later on memos generate new memos, reading literature generates memos, sorting and writing also generate memos—memoing is never done! Memos slow the analyst's pace, forcing him/her to reason through and verify categories and their integration and fit, relevance and work for the theory. In this way, he/she does not prematurely conclude the final theoretical framework and core variables. 
Comparative reasoning in memos—by constant comparisons—undoes preconceived notions, hypotheses, and scholarly baggage while at the same time constantly expanding and breaking the boundaries of current analyses. Memos are excellent source of directions for theoretical sampling—they point out gaps in existing analyses and possible new related directions for the emerging theory. Clearly the preconceived approach and framework of QDA research is in conflict with the freedom of memoing. The conflict is most often resolved by the preponderance of QDA research and GT loses this vital aspect. 
Throughout the constant comparative coding process, the researcher has been capturing the emergent ideation of substantive and theoretical categories in the form of memos. Once the researcher has achieved theoretical saturation of the categories, he/she proceeds to review, sort and integrate the numerous memos related to the core category, its properties and related categories. The sorted memos generate a theoretical outline, or conceptual framework, for the full articulation of the GT through an integrated set of hypotheses. 
Ideational memos are the fund of GT. Theoretical sorting of the memos is the key to formulating the theory for presentation or writing. Sorting is essential—it puts the fractured data back together. With GT, the outline for writing is simply an emergent product of the sorting of memos. There are no preconceived outlines. GT generates the outline through the sorting of memos by the sorting of the categories and properties in the memos into similarities, connections and conceptual orderings. This forces patterns that become the outline. 
To preconceive a theoretical outline is to risk logical elaboration. Instead, theoretical sorting forces the "nitty gritty" of making theoretically discrete discriminations as to where each idea fits in the emerging theory. Theoretical sorting is based on theoretical codes. The theoretical decision about the precise location of a particular memo—as the analyst sees similarities, connections and underlying uniformities—is based on the theoretical coding of the data that is grounding the idea. 
If the analyst omits sorting, the theory will be linear, thin and less than fully integrated. Rich, multi-relation, multivariate theory is generated through sorting. Without sorting, a theory lacks the internal integration of connections among many categories. With sorting, data and ideas are theoretically ordered. Sorting is conceptual sorting, not data sorting. Sorting provides theoretical completeness. Sorting generates more memos—often on higher conceptual levels—furthering and condensing the theory. It integrates the relevant literature into the theory, sorting it with the memos. 
Sorting also has a conceptual, zeroing-in capacity. The analyst soon sees where each concept fits and works, its relevance and how it will carry forward in the cumulative development of the theory. Sorting prevents over-conceptualization and pre-conceptualization, since these excesses fall away as analyst zeros in on the most parsimonious set of integrated concepts. Thus, sorting forces ideational discrimination between categories while relating them, integrating them and preventing their proliferation. The constant creativity of sorting memos prevents the use of computer sorting as used in QDA work. 
While theoretical coding establishes the relationship among variables, analytic rules guide the construction of the theory as it emerges. They guide the theoretical sorting and subsequent writing of the theory. Analytic rules detail operations, specify foci, delimit and select use of the data and concepts, act as reminders of what to do and keep track of and provide the necessary discipline for sticking to and keeping track of the central theme as the total theory is generated. 
There are several fundamental analytic rules. First, sorting can start anywhere. It will force its own beginning, middle, and end for writing. The important thing is to start. Trying conceptually to locate the first memos will force the analyst to start reasoning out the integration. Once started, analyst soon learns where ideas are likely to integrate best and sorting becomes generative and fun. Start with the core variable and then sort all other categories and properties only as they relate to the core variable. This rule forces focus, selectivity and delimiting of the analysis. Theoretical coding helps in deciding and in figuring out the meaning of the relation of a concept to the core variable. This theoretical code should be written and sorted into the appropriate pile with the substantive code. Once sorting on the core variable begins, the constant comparisons are likely to generate many new ideas, especially on theoretical codes for integrating the theory. Stop sorting and memo! Then, sort the memo into the integration. 
The analyst carries forward to subsequent sorts the use of each concept from the point of its introduction into the theory. The concept is illustrated only when it is first introduced to develop the imagery of its meaning. Thereafter, only the concept is used, not the illustration. All ideas must fit in somewhere in the outline or the integration must be changed or modified. This is essential for, if the analyst ignores this fitting of all categories, he/she will break out of the theory too soon and necessary ideas and relations will not be used. This rule is based on the assumption that the social world is integrated and the job of the analyst is to discover it. If he/she cannot find the integration, he/she must re-sort and re-integrate the concepts to fit better. The analyst moves back and forth between outline and ideas as he/she sorts forcing underlying patterns, integrations and multivariate relations between the concepts. The process is intensely generative, yielding many theoretical coding memos to be resorted into the outline. Again it cannot be done by the simple code and retrieve of computer sorting. 
Sorting forces the analyst to introduce an idea in one place and then establish its carry forward throughout the theory when it is necessary to use it again in other relations. When in doubt about a place to sort an idea, put it in that part of the outline where the first possibility of its use occurs, with a note to scrutinize and pass forward to the next possible place. Theoretical completeness implies theoretical coverage as far as the study can take the analyst. It requires that, in cutting off the study, he/she explains with the fewest possible concepts and with the greatest possible scope, as much variation as possible in the behavior and problem under study. The theory thus explains sufficiently how people continually resolve their main concern with concepts that fit, work, have relevance and are saturated. 
Always keep in mind that GT methodology is itself a GT that emerged from doing research on dying patients in 1967. It was discovered, not invented. It is a sure thing for researchers to cast their fate with. It was not thought up as a proffered approach to doing research based on conjectural "wisdoms" from science, positivism or naturalism. It is not a concoction based on logical "science" literature telling us how science is ought to be. 
GT gives the social psychological world a rhetoric—a jargon to be sure—but one backed up by systematic procedures. It is not an empty rhetoric, but unfortunately it often takes time for GT procedures to catch up to rhetoric with "grab." Part of the delayed learning is the remodeling—hence blocking—by QDA requirements, especially the accuracy quest. 
One promise is that the abstraction of GT from data—generating GT—does away with the problems of QDA that are "scientized" on and on. As the GT researcher (especially a PhD student) does GT analysis that produces a substantive, conceptual theory with general implications—not descriptive findings—he or she will advisably steer clear of the quicksand of the descriptive problems. QDA problems are numerous. A short list of these would include accuracy, interpretation, construction, meaning, positivistic canons and naturalistic canons of data collection and analysis of unit samples, starting with preconceived structured interviews right off, sequencing frameworks, preconceived professional problems, pet theoretical codes, etc and etc. The list is long, the idea is clear. 
"Minus mentorees" should be cautious, in their aloneness, about seeking too much guidance from "one book read" mentors and the intrusive erosion that results as these mentors try to make sense of GT in their QDA context. They should seek help from people who have written a GT book. 
The time for GT to explain and be applied to "what is going on" means leaving the onslaught of QDA methodologies, which so erode it and then remodel it. Evert GUMMESSON says it clearly in his recent paper, "Relationship marketing and the New Economy: it's time for De-Programming" (2002). What GUMMESSON says about marketing applies equally to nursing, medicine, education, social work and other practicing professions as well as academic work.
"Today's general textbooks perpetuate the established marketing management epic from the 1960s with the new just added as extras. It is further my contention that marketing education has taken an unfortunate direction and has crossed the fine line between education and brainwashing. The countdown of a painful—but revitalizing—process of deprogramming has to be initiated.
What do we need in such a situation? A shrink? No, it is less sophisticated than that. All we need is systematic application of common sense, both in academe and in corporations. We need to use our observational capacity in an inductive mode and allow it to receive the true story of life, search for patterns and build theory. Yes, theory. General marketing theory that helps us put events and activities into a context. This is all within the spirit of grounded theory, wide spread in sociology but little understood by marketers. My interpretation of a recent book on the subject by Glaser (2001) is as follows: 'take the elevator from the ground floor of raw substantive data and description to the penthouse of conceptualization and general theory. And do this without paying homage to the legacy of extant theory.' In doing this, complexity, fuzziness and ambiguity are received with cheers by the researchers and not shunned as unorderly and threatening as they are by quantitative researchers. Good theory is useful for scholars and practicing managers alike." (GUMMESSON, 2002, pp.585-586). 
I trust that this paper demonstrates how freedom from QDA requirements will allow unfettered GT procedures to result in generated theory that fulfills GUMMESSON's vision. 
Baker, Cynthia; Wuest, Judith & Stern, Phyllis (1992). Method Slurring, The Phenomenology/Grounded Theory Example, Journal of Advanced Nursing, 17, 1355-1360.
Creswell, John W. (1998). Qualitative Inquiry and Research Design. Thousand Oaks, CA: Sage.
Glaser, Barney G. (1978). Theoretical Sensitivity: Advances in the Methodology of Grounded Theory. Mill Valley, Ca.: Sociology Press.
Glaser, Barney G. (1992). Basics of Grounded Theory Analysis. Mill Valley, Ca.: Sociology Press.
Glaser, Barney G. (Ed.) (1993). Examples of Grounded Theory. A Reader. Mill Valley, Ca.: Sociology Press.
Glaser, Barney G. (Ed.) (1994). More Grounded Theory Methodology. A Reader. Mill Valley, Ca.: Sociology Press.
Glaser, Barney G. (Ed.) (1995). Grounded Theory 1984 to 1994. Mill Valley, Ca.: Sociology Press.
Glaser, Barney G. (1998a). Doing Grounded Theory. Issues and Discussions. Mill Valley, Ca.: Sociology Press.
Glaser, Barney G. with the assistance of W. Douglas Kaplan (Ed.) (1998b). Gerund Grounded Theory: The Basic Social Process Dissertation. Mill Valley, Ca.: Sociology Press.
Glaser, Barney G. (2001). The Grounded Theory Perspective: Conceptualization Contrasted with Description. Mill Valley, Ca.: Sociology Press.
Glaser, Barney G. & Strauss, Anselm L. (1965). Awareness of Dying. Chicago: Aldine Publishing Co.
Glaser, Barney G. & Strauss, Anselm L. (1967). Discovery of Grounded Theory. Mill Valley, Ca.: Sociology Press.
Gummesson, Evert (2002). Relationship Marketing and the New Economy: It's Time for De-Programming. Journal of Services Marketing, 16(7), 585-589.
Lowe, Andy (1997). Managing to Post Merger Aftermath-Default Remodeling, Dept of Marketing University of Strathclyde (Grounded Theory Review).
May, Katharyn A. (1994). Abstract Knowing. The Case for Magic in Method. In Janice Morse (Ed.), Critical Issues in Qualitative Research Methods (pp.10-22).Thousand Oaks, CA: Sage.
Morse, Janice (1994). "Emerging from the Data." Cognitive Processes of Analysis in Qualitative Research. In Janice Morse (Ed.), Critical Issues in Qualitative Research Methods (pp.23-41). Thousand Oaks, CA: Sage.
Morse, Janice (1995). Editorial. Qualitative Health Review, 5, 147-149.
Barney G. GLASER received his BA degree at Stanford 1952. He studied contemporary literature for a year at the Sorbonne, University of Paris, and he spent two years in the army one of which was in Freiburg, Germany, where he became fluent in German and studied literature at University of Freiburg during off-hours. He received his PhD from Columbia University in 1961. He then went to University of California San Francisco, where he joined Anselm STRAUSS in doing the dying in hospitals study and in teaching PhD students methods and analysis. He then published his dissertation as a book "Organizational Scientists: Their Professional Careers" and published over 20 articles on the subject of professional careers and the dying research. Then "Awareness of Dying" was written in 1965 and immediately became a resounding success. Upon request, GLASER and STRAUSS wrote the "Discovery of Grounded Theory" in 1967 to show how the dying research was done. Another big success! Then followed two more books on dying and one on chronic illness and one on status passage with STRAUSS. Since then GLASER has written 13 more books using and about grounded theory. Most are now published by Sociology Press. He has written countless articles. In 1998 he received an honorary doctorate from Stockholm University, Sweden. His works are read throughout the world.
Barney G. Glaser PhD, Hon PhD
Mill Valley, Ca 94942
Tel: 415 388 8431
Fax: 415 381 2254
Judith HOLTON is Research Development Consultant at Holland College, Charlottetown, PE, Canada. She holds a Bachelor of Arts (Dalhousie) and a Master of Arts (Royal Roads). She is currently engaged in doctoral research using grounded theory, as a PhD candidate at University College Northampton, UK.
Judith A. Holton
Research Development Consultant
140 Weymouth Street
Charlottetown, PE C1A 4Z1
Tel: 902 566 9607
Fax: 902 566 9629
Glaser, Barney G. with the assistance of Judith Holton (2004). Remodeling Grounded Theory [80 paragraphs]. Forum Qualitative Sozialforschung / Forum: Qualitative Social Research, 5(2), Art. 4, http://nbn-resolving.de/urn:nbn:de:0114-fqs040245.