Processing the Annual Review—2009

I want to write out some notes from my talk with my committee. Overall, my impression is that the actual researching (the doing of grounded theory) may be longer and harder than I envision. I got a better idea about the flow of doing the chapters, so I’ll summarize it below as a list:
• Come back to chapter 1 AFTER the research is complete
• Don’t wait until after research is complete to do the Lit Review section (this was something I had planned to wait to do)
• Chapter 3 may take longer than I anticipate because I may change things that I am doing
• Chapters 4 and 5 (findings and implications) will come more quickly after research is complete

We talked a fair amount about timeline and the flow of work. First drafts will flow through Rich until they are acceptable to go out to the whole committee. I expressed the desire to try to get an entire draft completed first and then work on revision. Now that I look at it, it seems to me that this goal may be wrong-headed. First, it puts a massive text in Rich’s lap, and second, moving chapter by chapter may be more realistic. Anyway, I have my first goal of getting a draft of the lit review done soon. I had the impression that the lit review is good for the entire committee. The most important thing about the timeline is that the dissertation needs to be all done at least one month before the defense (and preferably sooner than that). That means I should shoot for a January 1 completion data. Oh, my! I hope I have a productive summer.

We talked for quite a while about my research methodology. A big topic was the hybrid nature of my research where it is grounded in the traditional sense of analyzing slices of data, by hybrid in being able to go back to the massive database. One point that Rich brought up that I thought was interesting is about what a failed study would reveal. By failed he meant that what would be revealed if the findings from my by necessity narrow look at the database (I will actually be sampling a very small portion of what is available) and my findings were shown to be wrong or not to carry out? So much of our research in Composition is based upon making conclusions from small samples, but this result would indicate that this practice is perilous. We shall see. The big question is whether I will be able to take an emerged core category and query for that throughout the entire database. Right now, Fred says he doesn’t know how to query the text inside the database. The labor involved in taking out the separate texts and putting them into Atlas or Nvivo would be immense! Hmm… . I wonder about Nvivo. I believe it has a SQL database on the back end, so I wonder if I could import the entire database into Nvivo? Now that would make NVivo worth using. I can hardly imagine that it can do this large scale importing, but I’ll check it out.

Rich and I have talked about how the eventual methodology I will use will be noteworthy and worth an article when I’m all done. I have to admit that I don’t see it yet. I think it has to do with the back and forth of both sampling the database and then wide-scale querying of results inside the database. So it is how this database is used for research that is significant. That is what I am seeing at this point.

All and all I am beginning to see the enormity of the work to come like a large mountain to climb. I have to remember that mountains are climbed one step at a time. I think it helps to have a clear path, and though I feel pretty clear about my path, I need to clarify it more. And start walking!

Of first codings

I'm near the end of my first round of coding my first slice of data. I've been engaged in what is called micro-analysis, a line-by-line (even word by word) look at these textual artifacts. In this blog post, I want to write about my experience doing this micro-analysis (and not get into what I am seeing so far).

My initial task has been to underline sections of text and ask "What is going on here?" I'm looking at objects, events, acts, or happenings in the text. I've struggled mightily with the concept of categories—what is a category? That is what I am supposed to look for. Ian Dey specifically distinguishes them from names or labels. Gratefully, Strauss and Corbin have a helpful section on Open Coding. I looked at what I was doing, and I would say predominantly I was labeling what I saw. I had a lot of continuous tense verbs—showing evidence, affirming success, looking at draft, fitting in bounds. I see now that I was doing a lot of identifying and naming of what is happening for the students or happening in what they are doing.

The difficulty I have found is finding the right words for the labels. Often there are subtle shades of difference. What's the difference, for instance, between "showing evidence" and "offering specifics?" I'm using different language to describe the same essential thing. I saw a lot of where students would point to some fault or problem in either their draft of how they understood the assignment. Is it "admitting a fault" or "not fitting in bounds" or is there a better language to capture this form of self-assessment. But, as this last sentence illustrates, I have begun to categorize (I guess that is what I am doing) what is going on. I have grouped both pointing to faults or errors and point to successes as a form of self-assessment. Or should I call it "assessing the performance?" The problem of finding the right language to describe and label what I am seeing is difficult. It all seems a muddle, and it gets more muddled if I use different language. I've tried a few times to put aside the data and try to code the next set of artifacts with fresh eyes to see if I use different language. More muddle.

Let me describe two other things I have done as I did this first-level of micro-analysis. With most all the artifacts, I wrote a short note describing in a larger sense what was going on in this writer's review and what I thought of it. I also tried something Strauss and Corbin suggested, and that is to dig deeply into key words. For instance, I took the words "I believe" and "I realized" and wrote a string of synonyms beneath them trying to open up what the student may be really meaning when they say those words. This was helpful in places. I also noted what I considered "in vivo" names for what is going on. "Fit the assignment" is one example of what I marked as an in vivo term, though I have not decided to use it as a code yet.

Once I had coded all the artifacts, I took another blank sheet of paper and tried to determine classes of labels or what is going on. I was then able to fit a number of other labels within these classes. "Assessing the performance" became one where I could fit both self-assessment and peer assessment under that larger term. But things seem to be nested in loose ways, and how do I know when a sub-category should be its own category. For instance, I used the term "processing a problem" as a category which I had also put as a sub-category under "assessing performance." Both involved self and peer assessment. So where does this self and peer assessment go?

I'm used to from my literary analysis background to doing close reading, and I find myself slipping into this mode. I am not sure how to describe it, but at a certain point trying to process the muddle I feel like I have abandoned the identifying of categories, properties and dimensions.

I know that "constant comparative analysis" is a cornerstone of grounded theory, so I have attempted to do some comparing as I have reprocessed my initial micro-analysis. After I had spent some time just pulling together a comprehensive list of labels and groupings of labels (categories?), I decided to zero in on one common move I saw students doing and that was what appeared to be an end point generally described as "what I see now" or "what I realize" or "what I now know or know to do." Or even what I see to do next.

In writing a short memo about one of my artifacts, I saw that the student was telling a story, a narrative of sorts, of getting to this end point of what I labeled "knowing what to do." I'm not completely happy with the language of this term, but it works for now. Then I looked for this move or story in the different artifacts, finding it in pretty much each one. I diagrammed this narrative as a kind of flow chart. It seemed that there were different paths students took to get to this point of "knowing what to do." The diagram is messy at this point with some redundancies and I will next need to clean up this flow chart. What I am not sure of right now is whether I am on track by creating this diagram (is it an emerging theory) or am I going off-base. Drawing the chart along with examining the data made me see that this end point is really a middle point in some cases. Not only is there a sequence leading to this "knowing what to do" but there is a forward looking projected action in many cases, and these "what I will change" or do next are not all the same.

The goal of initial micro-analysis and open coding is to begin to identify categories and develop some emerging thoughts on theory (or what is going on here). I believe I have nearly accomplished this task with what I have done, but I'm uncertain of what to do next. How do I approach my next slice of data? I know that I need first to do some memoing to process my analysis so far to see where to go next. For me, especially, this reflective writing is very useful. I think I will do another round of micro-analysis on a set of writer's reviews done by students further along in their freshman comp sequence (1302 students) and see if similar patterns are evident or others are there.

I'm thinking right now that muddle is not all bad. I need to embrace the muddle and keep "muddling forward" and not get too stuck in figuring everything out at this point.

My Preconceptions

I'm going to take a stab at putting down my theoretical assumptions and beliefs BEFORE I begin the analysis of my data. Although Grounded Theory seems to have the requirement that the researcher NOT come to the data with any sort of preconceived theory, I think this requirement gets misinterpreted. First, it is impossible not to have preconceived notions. We all look at the world through our own terministic screens. Dey stresses that the belief in atheoretical observation is a myth. The importance is to be open to data and be aware of the theoretical biases that you already possess. So here it goes.

I thought I would start with some assumptions I expressed about reflection in 2004 in a paper I wrote that summer at the Central Texas Writing Project:

My Assumptions about Reflection

1) Writing reflectively is a learned skill.
2) Reflection helps students formulate and gain ownership of their own knowledge
3) Reflection plays a mediating role for learning from experience
4) Reflection plays a mediating role for learning from experience.
5) Deep reflection becomes "reflexive" (or transformational)
6) Reflection helps students formulate goals and solve problems as they compose.

Surprisingly, I still hold these assumptions to be true. Underneath all of this is a certain agency that I believe this act of reflection possesses. I have thought of various comparisons to describe this agency characteristic I believe reflection has: Superman's telephone booth,and a catalyst for a chemical reaction are the chief two. It just hit me that I didn't include something that indicates a slower growth process such as an oven to bake. Reflection is not an oven; it is something that is more immediate, or at least in its local effects. Implicit in all of this also is the notion of positive change. The most idealistic change or reaction is Mezirow's notions of transformational learning which Qualley picks up on and calls "reflexive." The contemplation, examination, and critique of one's assumptions for thought, belief, or action is said to create almost magically a significant change. This is the home run of reflection.

I also believe more modest, but still significant, things can happen through the "mindfulness" reflection promotes in students. Because the prompt asks students to consider and be more aware of certain things, those items may become more defined or real to the student.

Underneath all of these assumptions also is what we might consider a belief in the magical nature of language. A lot of reflective activities done in the 80s stressed the value of verbal reflection (Boud). Yet, I am focusing on written reflections. These two kinds of reflection seem to share the idea that students putting their thoughts and feelings into words (into language) gain something from that activity. Since our thoughts and perceptions of the world are formed to a large degree within the framework of language, using language is significant for developing this thought. I prefer written reflection because the student has more time to consider what they are writing, and then they have this document to look back on.

What other assumptions do I have?
I also have a more specific notion of this generative power of reflection for Composition that I believe links directly with "invention." For writer's I believe that the most significant concern of the writer's is the negotiation of their "rhetorical stance." That is, reflection helps students position themselves (their text and their thinking) in terms of the writing situation. Reflection provides the space to be "mindful" of the writing situation and all the unique factors that come into play. It is also the place where phronesis can be enacted--that is the flexible application of general rules to specific contexts or ill-structured situations. Reflection, then, becomes the pedagogical activity that reactivates the concerns of invention.

So I bring to my examination of writing and essays a whole lot of baggage about writing process, invention and pre-writing strategies, concepts of the writing situation, and what constitutes writing growth. I firmly believe in the importance of drafting and revision as a means for working on a piece of writing--it isn't a one shot deal. This notion of the "writing feedback loop" and drafting cycle is paired with the developmental nature of learning and knowledge (i.e. Kolb's experiential learning cycle). All of these assumptions are important to me.

I also have a few assumptions regarding causes of poor reflection or a lack of reflection. These predominantly revolve around four things: learning styles, developmental factors, knowledge, and conceptual frameworks of the task. I don't know, but it could be that some people are just not hardwired to think and learn in reflective ways. Being a reflective person myself, I can hardly imagine this kind of person (but we do have George Bush as an example). King and Kitchener as well as other intellectual development models assert that reflective thinking is a higher order level of thinking that comes with more maturity and development. I don't buy this idea that younger kids can't be reflective about what they do, but it is a significant idea and one I am unsure of. Some research also showed that how well students reflected depended to a great degree both on their knowledge (how can they be mindful of something they aren't even aware of or think in ways if they don't have the knowledge to think that way) and their conception or mental schema of the task (if they see the task as being about XYZ when it really is about ABC, then of course they will fail or flail).

For now, these are the key preconceptions I can think of. For me, these thoughts seem so natural and self-evident; thus, it is so important for me to get them down and in the open so I am aware of them. They, of course, are not natural at all.

Grounding Grounded Theory

I finally have finished Ian Dey's Grounding Grounded Theory, and I know that for the next few weeks (and perhaps the rest of my life) I will be processing this text. However, I want to take a big picture appraisal of his text.

He ends his book speaking about the growth of a misunderstanding about what grounded theory is that has been created by the proliferation of software tools for qualitative analysis. He is speaking from the perspective of 1999 looking back at what has happened in the 1990s. Pointing to work by Coffey in 1996, he describes how "the centrality of coding in both software for qualitative analysis and in grounded theory promotes an 'unnecessary close equation of grounded theory, coding, and software'" (qtd. in Dey 271). The mechanics of coding made more simple through computer software combines with the methodology of qualitative analysis that introduced the notion of "coding" to qualitative analysis to the point that "to code" meant "to engage in grounded theory." Any systematic analysis of data via "coding" meant grounded theory. And worse, as Dey notes, this convergence has resulted in "an uncritical attitude toward methodology" (271-2). Although these views equating qualitative analysis (coding) into one methodology (grounded theory) have been disputed, and I don't think anyone now would equate them, Dey speaks of a time when grounded theory became trendy and was used uncritically thanks to these computer software tools. Dey sums up the general problem in this trend: "It seems that anxieties over the convergence of qualitative research around a single methodology, which takes coding as the core of theorizing, may be well-founded" (272). The culprit seems to be the introduction of software tools that make coding and then the relating of categories through the retrieval of data easier. Dey's larger critique is that this process of theory generation and qualitative analysis becomes "mechanistic" and is done uncritically.

What his book reveals is that the we shouldn't blame the software tools for promoting a "mechanistic approach" because the seeds for this approach are there within grounded theory, as articulated by Glaser and Strauss. I'll list some of these tendencies in grounded theory that he says can lead toward a mechanistic and uncritical approach:

1. The inclination to consider coding as an aconceptual process
2. Observation is presented as atheoretical
3. Coding is said to be emergent rather than constructed
4. Theory is something we "discover"
5. Categories are conceived as separate concepts that are later connected
6. Process is analyzed through "slices of time" (rather than through an evolutionary analysis) (273)

He notes that it is fairly ironic and a case of unintended consequences that grounded theory has become reduced to these narrower, mechanistic approaches because of its stress on creativity and openness of analysis. As Dey states in his last lines of the book, "The dangers of a mechanistic approach cannot be avoided merely through exhortation. It requires reflection on the origins of the particular path taken and the problems it leaves unresolved" (273). Thanks to this book by Dey, I hope I will be able to avoid a mechanistic and uncritical engagement with my own research, and I believe he has given me tools to better understand what I will be doing as I start this analysis and the choices open to me and that I make.

The daunting task for me now seems to be to possess the adequate level of comprehension to do this research. It seems very complex. The thought occurred to me that I felt like a basketball player who has just been drafted to a new team, and he doesn't now the plays and the different schemes used for offence and defence. He's sent out on to the court to play, but things are happening so fast he doesn't know what is going on. Before I get out on the court (figuratively speaking), I hope to understand the dynamics at work more and about my choices. Since I am supposed to do a presentation for the May Workshop, I think I will make it on Grounded Theory so that I can use this task to get a stronger grasp on what I will be doing.

Preliminary Thinking on Sampling

My lunch today with Fred was primarily about how and what to sample in the TOPIC database. I've talked to Becky about this sampling question also, as well as John Horner (my "every man audience"). Plus, I'm doing a fair amount of thinking about it from my reading. What I want to do at this point is jot down some of my thoughts right now about what and how I would sample from TOPIC.

First, I need to keep in mind that my sampling is theoretical (that is, theory driven). This goal may be hard for the initial sample, and hard to predict for future samples because the emerging theory should guide the selection of data. I in no way need to make my sample in some way representative or of a certain number to be valid. No.

One thing I have noticed when others have thought about this question of what and how to pull data from the TOPIC vault is that people can see it as overwhelming. John Horner offered the advice to be careful not to set up a project that might take me years to do. He was seeing the vastness of the data and conjectured I would have to do some kind of research sample that would encompass the entire pool of data. No. No. Becky looked at it and paused at the amount and the complexity of the data. Each I felt had a sort of Grand Canyon moment--this is BIG.

Fred and I are in agreement that these Writer's Reviews documents should be viewed within the context of an entire Writing Cycle. In addition, he is leaning toward examining the relationship between what is said and happens in Writer's Reviews and what is said and done in subsequent drafts. Of course, this relationship is what is most interesting, and it is what other researchers have discovered is ambiguous. What students say they will do for revision and what they actually do can be quite different. I diverge from Fred a little bit in that I want to look at some Writer's Reviews just on their own as well.

The semesters of data I have chosen are these:
Year 04-05 1301 and 1302 (called (05-1301 and 05-1302)
Year 05-06 1301 and 1302 (called 06-1301 and 06-1302)

My current thought is that I will get a broad sample of writing reviews for my initial sample. I might also keep this sample fairly small so that I can make mistakes and it won't cost me a lot of time or effort, but it will be substantial enough for me to sink my teeth into the researching. For each course, I thought I would grab Writer's Reviews from Essay #2 and Essay #3.

I wish I could do tables in here, but here is my proposed selection.

For each essay in each class and year, I would grab two Writer's Review from draft #1 and two Writer's Reviews from draft #2 (4 Writer's Reviews for each essay and each year). That adds up to 32 isolated Writer's Reviews. In addition, I would pull a sample of full Essay Cycles from each course and each essay. A full Essay Cycle includes the every draft, every peer response, every Document Instructor response and grade, and every Writer's Review for the entire draft. I propose pulling two full Essay Cycles from each course and for each essay--that equates to 16 full Writing Cycles. I could modify this number down to 12 full Essay Cycles (6 from 1301 and 6 from 1302). 12 sounds more manageable to me, so I am not sure.

Fred might say only to do the full essay cycles, but I'm thinking I want the mix of Writing Reviews in isolation and then some in context. Hmm... I wonder if I should have some that are the same so that I could do a pass through of the data out of context and then look at it again in context. Hmm... . I have to consider that one.

This initial sample seems like it is large enough and broad enough to give me a base from which to then go in more particular directions depending up the emergence of my theory. My next quandary has to do with whether I will use a qualitative researching software tool. I probably will use one, but which one. It would be nice for me to be able to import a bunch of this text into the software program, but it looks like I may have to copy and past it in. ... More to look into.

On Coding in Grounded Theory

Dey has a fairly unsatisfactory chapter on coding because he truly shoots some holes into Grounded Theory in his discussion about coding. He opens with a good distinction between categorization and coding: "With categories we impute meanings, with coding we compute them. The former involves a creative leap, for 'comprehending experience via metaphor is one of the great imaginative triumphs of the human mind' (Lakoff, 1987, p. 303). The latter involves reduction and ready reckoning" (95). It is interesting how Dey uses uncommon words such as "impute" and "reckon"--one involves a leap, while the other involves reduction (pulling back and being conservative). These seem like contrary impulses.

Dey provides a bit of historical perspective on the term "coding" and notes with some ironic amusement that it has been the combination of grounded theory and computer analysis tools that has led to the term"coding" as being the term describing the key process in qualitative analysis. What is ironic is that coding comes from the quantitative use of survey methods where "coding" happens to analyze surveys that have "precoded" questions. The survey already has within it the concepts and categories so that when the surveys are analyzed the researcher identifies and assigns the appropriate codes to responses and tabulates them. Dey notes that in qualitative analysis the researcher does not have the conceptualization already complete as they review the data because it is yet to be accomplished (96).

However, Glaser and Strauss strongly believe "coding and analysis proceed jointly in grounded theory" (96). Thus they contrast grounded theory from other qualitative methodologies that would code first then analyze second (separating the processes). What this joint coding and analysis exactly looks like, I am not sure. In this section distinguishing coding in grounded theory, Dey notes a point Glaser and Strauss emphatically make: "they reject the method of coding data 'into crudely quantifiable form' in order to test hypotheses, since they are interested in generating theory rather than verifying it" (96). My first reading of this emphatic point was that they rejected the idea of analyzing to determine a code from which you next test (which is what my research design provisionally contains). I now see that I may have been reading too much into this quote. What they seem to say is not to count your data--there is no need to determine the number of occurrences of a concept or interaction, especially if this counting is done in order to support some hypothesis or theory the researcher has as they come to the data. As Dey says, "coding is governed only by theoretical relevance and is not concerned with the accumulation of supporting evidence" (96). I wonder about this claim since I believe computer analysis tools can easily provide you a graphical representation of the prevalence of a particular code (such as a tag cloud). I'll see this possibility more as I get into using these computer analysis tools. The last key points about coding Dey stresses is that coding is a method to make conceptualization explicit, and its function is "to generate rather than test a theory" (97). One must resist, then, impulses to turn coding into a hunt for verification, it seems.

The orthodox Grounded Theory view toward coding is that it proceeds in phases:

1. Categorize the data (open coding)
2. Connect the categories (theoretical or axiel coding)
3. Focus on a core category (selective coding) (98)
   

In his discussion about this breaking down of the data, especially in the initial open coding, Dey indulges himself by countering analysis with "holism." His basic point seems to be twofold--taking a big picture, holistic view may reveal things that narrow analysis will miss, and that through this holism we may tap into what he calls "direct understandings" of the world better. These direct understandings come from our "bodily experience in the world" and they may be "both rich and routine" (102). Dey seems to be going off on a strange tangent here, but I think his call to maintain a big picture awareness as one focuses on the microscopic is wise, and it is our "sensitizing experience" that can guide this holistic analysis.

Dey offers another innovation (we might say) toward coding. This time he offers a counter viewpoint to the notion of phases and the idea that you categorize first and then connect these categories. Dey's point is this: "categories cannot be considered in isolation. Categories acquire their meaning in part from their place in the wider scheme of things... . ...discrimination among objects may depend on their place in a larger taxonomy" (105). Rather than category sets, Dey chooses the metaphor of "category strings" to represent how categories exist within a network of other categories. Thinking of Lakoff, Dey is stressing the point that any category/categorization activates a larger conceptual framework, and we need to be aware of this network of connections. His point is that "in grounded theory, the division between open coding and axial coding needs to be treated with caution" (105). His point is that we need not wait to uncover links between categories as we open code (I think), and that we need to be aware of these larger "strings" of relationships within the categories we declare. It makes me think a bit about activity theory.

Next Dey considers Axial Coding in more detail, first discussing Strauss and Corbin's "coding paradigm" (1987). Strauss' coding paradigm examines: conditions, interaction among the actors, strategies and tactics, and consequences (106). The value of this coding paradigm is its clarity that makes the entire process of coding more manageable. You know what you are doing with this paradigm. Dey questions why this paradigm should be privileged? Glaser criticized Strauss' "coding paradigm" because it ignored Glaser's work on "theoretical coding": "Instead of 'forcing' the data to fit a pregiven paradigm, Glaser suggests we consider a range of theoretical options of which the proposed paradigm is only one" (107). Glaser in is 1978 book lists sixteen "coding families" that provide a range of options for coding (107). Glaser stresses that a coding family should only be used once it is indicated by the data. Dey questions what to do if more than one family could fit the data, thus making the choice of a coding family arbitrary.

The final part of this chapter discusses "core categories" in detail. Core categories are central for grounded theory. As Glaser believes, "The aim of producing a grounded theory that is relevant and workable to practitioners requires conceptual delimitation" (110). Core categories are where the researcher deliminits their categorization. Glaser believes a core category has to "earn its privileged position" (111) by containing these qualities (or meeting these criteria): "[it] has to be central, stable, complex integrative, incisive, powerful, and highly variable" (111). Dey questions why only ONE category is chosen, and not more. He blasts grounded theory at this point for forcing a fit, and what obviously involves an "elimination of alternative accounts" (112). He criticizes the core category as also being paradoxical in terms of its role as both dependent and independent variable.

All these critiques come back to the basic difficulty of dealing with subjectivities when coding data. This quote from Dey seems to express the difficulty well: "The construction of a category or the appropriateness of assigning it to some part of the date will undoubtedly reflect our wider comprehensions--both of the data and what we are trying to do with it. The researcher (who brings to categorization an evolving set of assumptions, biases, and sensitivities) cannot be eliminated from this process" (104). I knew this truth before, but seeing it spelled out in more detail in this chapter makes the entire process seem more and more daunting.

Categories and Categorization in Grounded Theory

This blog post will attempt to make sense of Ian Dey's long chapter on Categories and Categorization in Grounded Theory. Categories are maddeningly confusing, and at times it seems what Dey reveals in this chapter is like the soft underbelly of a dragon.

Let's start by presenting definitions:
Categories are conceptual "and never just a name or a label" (49). Categories are said to stand alone and refer to a class of things.

Properties, though, can't stand alone and are "conceptual characteristics of a category" (51). They refer to external to external relationships and relate through interaction.

Dimensions represent the spectrum of variation possible within properties. Dey says, "Identifying dimensions therefore involves (internal) differentiation rather than (external) comparison" (52).

I will chart out Dey's example illustrating these three concepts below:

Category-- color
Properties-- shade, intensity, and hue
Dimensions--intensity can be high or low, hue can be dark or light

Dey says this example illustrates the orthodox distinction between properties and dimensions. Here is one good quote from Dey: "Whereas properties and dimensions 'belong' to the thing itself, the categories to which we assign it do not belong to the thing itself but are part of how we choose to classify it" (54). He stresses the point that categories are derived through comparison.

Dey uncovers a confusion within the analytic processes that each of these three refers to. He says that we can apply all of these analytic processes to the same phenomenon. He stresses that each process of analysis has a different purpose: "We use categories to distinguish and compare; we identify properties and attributes to analyze agency and effects; and we measure dimensions to identify more precisely the characteristics of what we are studying" (57). Dey feels that distinguishing these three concepts through purpose is better than seeing them as varying levels of abstraction.

The next section in the chapter presents the classical view of categories as elements in theory. Dey presents this summary before undercutting it completely. Theorizing involves discovering how categories relate to each other, and GT seems to have two ways of relating: one is through relations of similarity and difference, and second through connection and interaction. He provides this example of relating a cat, dog, and bone.

Formal relations based on similarity and difference: puts the cat and dog together
Substantive relations based on connections: puts the dog and bone together

Our understanding of substantive connections is based upon our observation of the process.

Digging deeper, Dey tackles Glaser's "concept indicator model" which Glaser claims "provides the essential link between data and concept" (qtd. in Dey 60). The meaning of the category (or code) is defined in terms of its indicators. Dey uses the example of prejudice as a concept (category). We can't observe the abstract concept of prejudice, but we see it in action, so to speak--through its "indicators." We can look at statements or actions and identify them as indicators of prejudice. Glaser believes that constant comparative analysis slowly builds concepts through "the careful combination of indicators." Concepts, then, become "the 'sum' of its indicators (61). Dey has some questions about Glasers concept-indicator model.

Summing up the chapter, Dey points out that one of the special characteristics of grounded theory "is its firm location in an interactionist methodology." It is focused on explicating social processes in dynamic terms. I think this characteristic of GT is important to remember. The elements of theory--categories, properties, dimensions--and the process of categorization--constant comparison, focus on indicators--all facilitate this interactionist methodology.

Next Dey digs the deepest into "categorization"--the fundamental process of distinguishing and classifying. Here is where things get messy. Dey explores modern developments in categorization that question the simple concept-indicator model of Glaser and it process of basing categorization upon judgments of similarity and difference that seems to figure so largely in GT. As Dey says quite simply: "The identification of categories on the basis of similarity and difference turns out to be rather problematic ... [and] in practice the process of drawing distinctions is much more complicated and ambiguous than the concept-indicator model allows" (66). Great. Pull the rug out from underneath me. Dey reveals that categorization is much more variable than Glaser describes it in his concept-indicator model and "it challenges any simple assumption that categories are 'indicated' by data in a straightforward way" (75).

Dey goes on to describe three alternative understandings of categorization from scholarship done since the 1967 advent of GT. The chapter is dense, so I will include Dey's own summary:

"In the above discussion, we can identify at least four different accounts of categorization. First, we have the classic account, which assumes that category boundaries are crisp, membership is based on common features, and relations between categories are governed by logical operations. Second, we have 'fuzzy' sets, where category boundaries become vague, membership is graded, and relationships between categories become a matter of degree. Third, we have the 'prototypical' model, which stresses the role of category exemplars and shifts focus from membership to degrees of fit. Finally, we have categorization in terms of 'idealized cognitive models' (this is Lakoff) which 'motivate' the creation of categories through various forms of 'chaining' and 'extension'" (86).

After revealing the basic instability of categories and the process of categorization, Dey seems to mollify his reader by stating a hopeful message: "while the processes of categorization may not be strictly logical, neither are they entirely arbitrary" (87). He then provides a number of things the researcher needs to do to render her analysis not entirely arbitrary:

1. Render the cognitive processes of categorization explicit (i.e which of the four approaches to categorization will you take)
2. Assess the adequacy of the cognitive process in terms of the underlying cognitive assumptions employed.
3. Recognize the various processes involved in categorization
4. Identify the aims of categorization (for example, prediction or inference)
5. Make more explicit the grounds (cue or category validity) on which these categories can be realized
6. Identify the underlying conceptual models and make explicit their metonomyic or metaphorical extensions (a la Lakoff) (87)
   

This list is quite daunting, and to me seems to require a level of self-awareness that might be impossible to achieve. I believe that I can attempt to define these foundations for my categorization before I start coding, but I may not be able to get far until I start coding data. Coming back to this list will probably be an important thing to do after my pilot study.

The larger point of Dey's chapter is summed up in a statement he makes near the end of the chapter: "In grounded theory innocence is preserved and bias precluded by allowing categories to emerge from (and hence correspond to) the data. But Lakoff's analysis suggests that such innocence is impossible to achieve. We think in terms of categories and our categories are structured in terms of our prior experience and knowledge" (92-3). Grounded theory, Dey, believes must reassess how it categorizes in light of new theories that challenge and expand "how categories are actually assigned and used in the production of knowledge" (91).

I always knew that the original description of generating categories from the data was naive; however, Dey has overwhelmed me with the detailed description of this inadequacy. Yet I would rather be aware of these problems and strategically (and perhaps rhetorically) chart my approach to categorization and analysis that enter this forest without a plan. I am hopeful that Dey will offer more explicit suggestions as I keep reading.