Tuesday, November 18, 2008

Sensitizing Categories

Entry for 17 November 2008:

Here is another little methods piece I did for the University’s research methods website, which I think is of more general interest.

Sensitizing categories, as proposed by Glaser & Strauss (1967), are concepts brought by researchers to aid qualitative analyses, but which do not unduly affect the results the emerge from the data. In addition to being potentially useful tools, the concept of whether or not to consciously employ sensitizing categories in one’s research raises important general issues for qualitative researchers to consider.

1. Before starting a qualitative study, should I avoid reading the literature, in order to keep an open mind?

In 1980’s, at the beginning of the qualitative revolution in social science research, many researchers were convinced that the only way to escape the narrowness of previous quantitative social science research was to avoid reading about your topic until after collecting and analyzing your data. In this way, by keeping your mind in a pristine condition, you would be able to maintain an unbiased view of the data. You wouldn’t have to “bracket” your assumptions (a favourite metaphor of phenomenologists), because you wouldn’t have any assumptions.

This view is attributed to Glaser and Strauss (1967), founders of the Grounded Theory approach to qualitative research, and Glaser (1978; as opposed to Strauss, 1998) does in fact advocate avoiding doing pre-research literature reviews. This has a certain appeal, and also justifies jumping right into one’s research without lengthy, boring sitting around in libraries reading books and articles. You could go directly, “back to the things themselves” (to quote the famous dictum of the phenomenologist Husserl, 1913/1982).
Questions for reflection:
•Why is this no longer the dominant view of qualitative researchers?
•Whose interests might be served by a return to a more traditional view of “literature review first”?
•Under what conditions might “literature-free” research be justified?
•How can qualitative research retain its openness and freedom, while still grounding itself in the relevant scientific literature?

The inevitability of “bias”

Whether the belief in assumption-free qualitative research can be best understand as misplaced optimism about the openness of the human mind, as an expression of laziness, or as dangerous naivet√© leading to unconscious, unchecked bias, it is not the dominant view of qualitative researchers today. In fact, it’s not clear that this “eternal sunshine of the spotless mind” was ever really subscribed by more than a few qualitative zealots. For example, an important early text (Bogdan & Taylor, 1975) urges readers preparing to do qualitative observational studies to read the literature in preparation for entering the field, so as to avoid committing faux pas with informants. Clearly, it is necessary to know something about what you are studying before starting to collect data; the question is, what and how much? There is also the question of how to use what one does know.

Philosophers since Kant have generally understood that it is impossible to know anything without bringing some prior knowledge, or forestructure, along. Not only is it impossible not to drag one’s prior knowledge along, but this prior knowledge is what makes the knowing possible. Without prior assumptions, categories, and words, we would be faced with what the American psychologist-philosopher William James (1890) referred to as a “buzzing, blooming confusion”, in other words, chaos. We wouldn’t even know there was something there to be known; everything would sound like noise, that is, meaningless, random stuff. For example, even something as seemingly straightforwardly obvious as an impression of the colour red is only made possible by the fact that our sensory system is biologically tuned to certain light frequencies and culturally programmed to divide the light spectrum into certain major categories. Awareness of these prior knowledge structures has led to whole fields of study into the nature and content of the human knowledge system, including perceptual and cognitive psychology, depth psychology (psychoanalytic and Jungian), cultural anthropology, and ethnomethology (Garfinkel, 1967), among others.

Even worse, failure to acknowledge the role, influence, or aid of prior knowledge means that we cannot therefore know the extent to which we are being limited or influenced by that knowledge, thus blinding us to our biases, whether these are biological, psychological or cultural in origin.
Exercise: Take a blank piece of printer paper. Fold it half the short way, top to bottom. Run your finger along the fold, looking carefully at how the paper has taken the fold. Now unfold the paper, and fold it the long way, right to left. Once again, take a close look at the fold, at how bumpy or not the fold is. Note that that one of the two folds you made is much more smooth while the other is more bumpy. The bumpy fold was done against or across what is referred to as the bias or grain of the paper, while the smooth fold was done with the bias or grain.
Finally, find a good dictionary and look up the meaning and origin of the word “bias”. Why is there bias in paper? What other things have bias? Is it possible or even desirable to make “unbiased” paper? What implications does this parable have for research in general, and qualitative research in particular? What does this have to do with the concept of “sensitizing categories”?

Using Sensitizing Categories

In retrospect, Glaser’s (1978) argument against pre-research literature reviews appears to have been misunderstood. First, he doesn’t say you should read nothing. Rather, his view is that the researcher should read books on qualitative method, especially his books, to help you develop an open, creative approach to your data. He also says you shouldn’t record your interviews or talk to anybody about your emerging theory (Wikipedia, 2008); fortunately, nobody today listens to those methodological guidelines either! Second, the reason Glaser says it’s a bad idea to read the research literature before doing your research is that it will desensitize you, that is, make it more difficult for you to hear what your data are trying to tell you. He doesn’t want you to read right past your data, saying to yourself, “Oh, of course that’s an example of X”, rather than stopping to really listen to what a participant is trying to tell you. So the real point is that anything that interferes with hearing your data with a fresh, curious ear is bad and should be minimized.

But Glaser (1978) also argues in favour of “theoretical sensitivity”, the use of sensitizing concepts that help you hear what is going on in your data. How is this possible?

Glaser (1978) and Strauss (Strauss & Corbin, 1998) famously disagreed on how to do this. While Glaser favours a more open-ended, creative, unstructured use of prior concepts, Strauss and Corbin put forward a “paradigm” of organizing categories, which is the basis of what they refer to as “axial coding” (i.e., organizing your data within a set of broad headings or domains). Exactly what this means is subject to varying interpretation. On the one hand, it can be understood as a general social theory that can be applied to a wide range of social phenomena, as Strauss & Corbin appear to imply. On the other hand, others (e.g., Elliott & Timulak, 2005) argue that the axial coding or domain structure of a topic emerges over the course of analyzing the first 3 to 5 interview protocols, and reflecting the interaction between the researcher’s interests and what participants are interested in telling about the topic. For example, it is appears to be common for informants to want to say much more about the background or context of the phenomenon than the researcher is interested in studying(e.g., Rhodes et al., 1994).

Sensitizing categories as research topic domains. In this latter view, then, the most useful sensitizing categories are broad headings that remind the researcher to listen for particular aspects of experiences, without constraining the specific nature of those aspects. For example, it is generally useful to organize one’s topic into a set of domains or subtopics such as:
• Context: What led up the targeted experience (e.g., realizing one’s ordinary way of seeing something was wrong; struggling and feeling confused; considering different possibilities; etc)
• Experience: What happened during the targeted experience (e.g., understanding a scientific concept such as gravity)
• Effects: What happened afterward, or flowed out of the targeted experience (e.g., feeling clarity and relief; applying the concept to other situations)
While such a framework of sensitizing categories or research domains seems obvious and straightforward, it often turns out need further adaptation. For example, participants may have little to say about the effects of an experience, but may make an important distinction between the immediate context or situation in which the experience occurs and its deeper context or background earlier in person’s studies or life. Thus, it seems important for the researcher to hold these sensitizing categories lightly and to be ready to adapt them to fit the particular research topic and data.

At the same time, even a primitive set of sensitizing domains will alert the researcher to listen for a broad range of aspects relevant to their topic. It will tell them where to look, but not what they should see there. Sensitizing categories organize the phenomenon being studied, without constraining it. Organizing the data into domains reduces the complexity of the analysis into more workable subsets of data that can be handled one at a time. This means that the researcher is less likely to be overwhelmed by the “buzzing, blooming confusion” of their data.
Exercise: Spend half an hour thinking about a research topic you are interested in. What specifically do you want to know about it? What are its aspects or phases? Is it something that happens in a single moment, or does it unfold gradually over time? Does it have a beginning, middle and an end? See if you can produce an initial set of headings to organize your topic. Try to write one or two interview questions for each heading.


Bogdan, R., & Taylor, S.J. (1975). Introduction to qualitative research methods. New York: Wiley.

Elliott, R. & Timulak, L. (2005). Descriptive and interpretive approaches to qualitative research. In J. Miles & P. Gilbert (eds.), A Handbook of Research Methods in Clinical and Health Psychology (147-159). Oxford, UK: Oxford University Press.

Garfinkel, H. (1967). Studies in ethnomethodology. Englewood Cliffs, NJ: Prentice-Hall.

Glaser, B. G. (1978). Theoretical sensitivity. Mill Valley, CA: Sociology Press.

Glaser, B. G. & Strauss, A. (1967). The discovery of grounded theory: Strategies for qualitative research. Chicago: Aldine.

Husserl, E. (1913/1982). Ideas pertaining to a pure phenomenology and to a phenomenological philosophy: General introduction to a pure phenomenology (trans. F. Kersten). Boston: Kluwer.

James, W. (1890/1981). Principles of psychology. Cambridge, MA: Harvard University Press, 1981.

Rhodes, R.H., Hill, C.E., Thompson, B.J., & Elliott, R. (1994). Client retrospective recall of resolved and unresolved misunderstanding events. Journal of Counseling Psychology, 41, 473-483.

Wikipedia. Grounded theory (Glaser). Retrieved on 12 Nov 2008 at: http://en.wikipedia.org/wiki/Grounded_theory_(Glaser)

No comments: