Triangulation and Disconfirmation

Triangulation and confirmation

Methodological triangulation is the use of different kinds of method in a single investigation. In the nursing literature, there was a debate over triangulation in the nineteen eighties and nineties. While less febrile than it was, the debate continues today. The crux of the debate is whether both “qualitative” and “quantitative” methods could support a single result (theory). Those who argued against confirming triangulation thought that different methods could be only complementary, each supporting a distinct and independent part of a study. In “A New Foundation for Methodological Triangulation” (Journal of Nursing Scholarship, 34: 269-275, 2002) and “Methodological Triangulation in Nursing Research” (Philosophy of the Social Sciences, 313: 40-59, 2001), Sandra Dunbar, Margret Moloney, and I argued in favor of the use of qualitative and quantitative methods to support a single study. (See my web page for links to these and other essays.)

In our discussion, we noted that results of the different methods might conflict, but we did not pursue the further questions that such conflict raises. How are investigators to handle inconsistent results from different methods? Can the results of a quantitative study show that a qualitative result is incorrect, or vice versa? Does one kind of method trump the other? Published literature normally touts successes, not failures. Hence, it is difficult to find concrete examples that help us think about how one kind of method might undermine or disconfirm results obtained by another.

Heat wave

An interesting example for reflecting on triangulation comes from the sociological literature on the 1995 Chicago heat wave. 739 people died in a week of record heat and humidity. In Heat Wave: A Social Autopsy of Disaster in Chicago (University of Chicago Press, 2002), Eric Klinenberg argued that differences in death rates among neighborhoods were explained by their differing “social ecologies.” Death rates were higher in North Lawndale, a poor and predominantly African-American neighborhood, than in South Lawndale, a Latino neighborhood. Analysis by the CDC showed that throughout Chicago, being elderly and living alone were risk factors for heat-related deaths. Klinenberg appeals to factors like crime rates and population density as contributing to social isolation of the elderly, which would make them vulnerable.

The data for Klinenberg’s research were primarily demographic, and would count as “quantitative,” if we apply the category as used in the nursing literature. An interesting challenge to Klinenberg’s work comes from Mitchell Duneier, who conducted a series of interviews with residents of North Lawndale and nearby neighborhoods (“Ethnography, the Ecological Fallacy, and the 1995 Chicago Heat Wave” American Sociological Review, 2006). 16 people died in North Lawndale. Interviewing neighbors and family members, Duneier found that of the 16, “at least 12 were living with families and had domestic relationships” (p. 682). 14 of the 16 were said to have gone out of the house regularly. Duneier’s interviews thus directly undermine Klinenberg’s explanation for the heat wave deaths in North Lawndale.

Text and numbers in conflict

Duneier’s work shows one of the ways in which interview results can disconfirm survey results. Duneier roughly conforms to a Popperian logic: he identifies a consequence of Klinenberg’s explanation (that many or most of the deceased in North Lawndale were socially isolated), and shows it to be false. Characterizing the logic of Duneier’s argument in this way, however, effaces much of its texture.

Duneier describes his method as “ethnographic,” but the word is used differently her than in the nursing literature (or the anthropological literature, for that matter). What he means is that his method involved “shoe leather.” He and a photographer walked the neighborhood, knocked on doors, and struck up conversations with passers-by. Had he not done this, e.g. had he made a random selection from the telephone book, it would have been very unlikely that he would have gathered the kind of information he did. Moreover, Duneier did not obtain his result by simple counting. His open, unstructured interviews were a crucial part of unfolding the stories that surrounded each unique individual. Sifting through these, he identified the themes that were key to understanding the ways in which the deceased may or may not have been made vulnerable by the social ecology of their neighborhoods.

Qualitative and quantitative methods in conflict

One of the reasons why the triangulation issue continues to be discussed is that qualitative researchers often feel threatened by quantitative work. Statistical methods have the status associated with big science, and the politics of qualitative and quantitative research continues to plague the discipline of nursing. Duneier’s work shows how there is less distance among methods than is commonly assumed. Methods associated with the qualitative research do not require a methodology that is opposed to quantitative research. Because the methods are consistent, a careful and imaginative researcher like Duneier can bring them into productive conflict, providing a richer and more empirically adequate picture.

Reductionism and Holism: the Practical Holism of Nursing Practice (Part 3)

Holism and Reductionism in Nursing

In two previous posts, I have explored some of the meanings of “reductionism” and “holism” found in the philosophical literature (see Reductionism and Holism: Three Varieties (Part 1) and Reductionism and Holism: Ontological (Part 2)). While these conceptions of holism and reductionism have been influential in the nursing literatures, they do not exhaust the uses of the terms “holism” and “reductionism.” Indeed, there is a use of “holism” in nursing that seems to fit neither the epistemological nor the ontological senses of the term. It is nicely illustrated in the following passage from Rosemary Ellis’s “Characteristics of Significant Theories” (Nursing Research 17 (3): 217-222):

Holism, if used as the appropriate view for aiding a patient, requires that one be concerned with any factor, be it physiological, social or any other, which affects the patient’s health. It requires that the factors be treated in combination, not in isolation. It also means that the combination is not the same as the sum over each factor. Nursing requires the recognition of the inseparability and interdependence of many factors. (p. 218)

What conception of holism is embedded in this text?

Practical Holism

The first point to notice is that the topic—that about which one is to be holist or reductionist—is not theory; it is practice. This means that Ellis is not expressing epistemological holism. That is, she is not taking position on how theories are related (e.g. whether sociology might reduce to biology, for instance). Nor does she seem to be taking a position on what exists (at least not directly, more on this below). Holism, according to Ellis, is something that might be “the appropriate view for aiding a patient,” and it requires a specific kind of concern. It is a practical commitment to treat the factors influencing a patient’s health in combination.
The “reductionism” that Ellis implicitly opposes is found in medicine. Physicians were portrayed in this period (and often since) as narrowly concerned with the patient’s disease or dysfunction. (Whether this is an accurate portrayal is an open question.) The nurse, by contrast, was charged with a broad responsibility for the patient and his or her environment. A reductionist approach in this domain would consider a patient’s disease or dysfunction in abstraction from the other factors. Because of her role in health care, a nurse cannot be so limited.

The broad responsibility of a nurse is the source of the idea that nursing is committed to holism. The nursing role demands that a nurse attend to any and all factors that affect a patient’s health. To limit nursing concern to one organ system, or to one dimension of psychological or social dysfunction, would be to ignore the broader context of patient health which has been central to nursing since Nightingale’s time.

Practical holism, then, is the commitment to address the whole of a patient’s health. Each of the many factors that affect a person’s wellness needs attention, and they must be attended to in a way that recognizes the interactions among them.

Ontological implications

Practical holism is different from epistemological and ontological holism. But what are its implications? Does practical holism entail strong (or weak) forms of ontological or epistemological holism?

Arguably, practical holism is inconsistent with strong forms of ontological holism. A strong ontological holist about minds, for instance, holds that the existence of minds is independent from the existence of bodies. The mind and body are causally linked, to be sure, but the connections are contingent. Changes in the nervous system need have no affect on the mind. Ironically, ontological holism about the mind supports a kind of reductionism in the practical realm. An intervention might target the mind alone without consideration of the body, for after all, they are independent.

In the nursing literature there are some very strong commitments to ontological holism – in the work of Rogers or Parse, for instance. If the foregoing arguments are correct, then the metaphysical views of Rogers or Parse are incompatible with the nursing commitment to address the whole patient.

Practical holism, Ellis says, “means that the combination is not the same as the sum over each factor.” This suggests that the mind and body (to continue the example) should be understood as interactive. They are not two ontological kinds linked by contingent inputs and outputs, rather, they are aspects of the same reality. The moderate holism of the gate control theory of pain (see Reductionism and Holism: Ontological for a brief discussion) is much better suited to the commitments of practical holism. Nursing interventions need to consider the physiological, psychological, and social dimensions of pain in combination. If these “factors” are conceptualized in a unified way—pain is both something experienced and a physical phenomenon—it will be natural to design interventions that are not merely a “sum over each factor.”

Conclusion: Nursing and the Three Varieties of Holism

Nurse scholars often characterize their discipline by a commitment to holism. This commitment is sometimes used as a basis for rejecting certain kinds of research as not properly part of nursing science. Causal models of biological, psychological, or social systems are often marginalized for this reason. The foregoing discussion provides a new perspective on such claims.

A commitment to practical holism is a consequence of a traditional conception of the nursing role. If nurses are charged with attending to the full gambit of patient needs, then nurses are committed to the whole patient. They could not be reductionist in the sense of attending to just one factor affecting health. Practical holism, then, is arguably central to nursing practice. Epistemological and ontological holisms are not similarly implied by nursing practice, and therefore they are not central to the nursing discipline.

It is plausible to reject causal models because they conflict with the holism only if one conflates the different varieties. Once the varieties are distinguished, we can appreciate some of the subtle relationships among them. The moderate ontological holism of the gate control theory of pain, for example, is more compatible with the practical holism of nursing than the strong holism of Rogers or Parse. Far from being marginal to the nursing discipline, some kinds of mechanistic causal modeling might be central to it.

Reductionism and Holism: Ontological (Part 2)

The story so far…

In Varieites of Reductionism: Three Varieties (Part 1), I discussed the problem of reductionism as it appears in nursing and described “epistemological reductionism.” This involves the relationship of one theory to another. There are two other kinds relevant to the issue of reductionism: ontological reductionism and practical reductionism. This post will articulate ontological reductionism.

Ontological reductionism and holism

Questions of ontology concern what exists and what kinds of things there are. Whether God exists, whether the number 2 exists, and whether cats have minds like ours are ontological questions. Questions of ontological reductionism ask whether one (purported) kind of thing can be identified with another. Thus, to say that there are no minds or spirits, only brains, is to take a reductionist position.

Note the difference between ontological and epistemological reductionism. In epistemological reductionism, the question about minds and brains is whether our knowledge of minds (and how they work) could be reduced to knowledge of brains (and how they work). The ontological question is whether there are two kinds of things (minds and brains) or just one (brains). The denial of ontological reductionism about the mind would hold that minds exist independently of brains.

Ontological holisms and reductionisms come in weak and strong forms. Strong holism contends that the two domains exist independently. Those who hold that the mind (or spirit) is a non-material entity which can survive the death of the body affirm a strong form of holism about the mind. Those who affirm that only brains exist affirm a strong form of reductionism.

Pain and weak holism

Middle positions are possible, and the Gate Control Theory of Pain provides an interesting example. The Gate Control Theory holds that pain signals travel along two kinds of neural circuit. One runs directly from the injured site to the motivational parts of the brain. The other goes through the higher processing, cognitive parts of the brain. The latter forms a “gate” which can block or modulate pain sensations. Is this a form of ontological reductionism? It seems so, insofar as pain is identified with neurological circuits. Notice, however, that the feeling of pain is not eliminated or explained away. It would be odd to be skeptical about the existence of the feeling of pain in a way that it is not odd to be skeptical about the existence of, say, souls.

The Gate Control Theory assumes that people feel more or less pain, depending on their emotional or cognitive state. Both the feeling of pain and the neurological states are assumed to exist. Hence, it is not strongly reductionist, since it does not deny that pains exist. Nor is it strongly holist, since the feeling of pain is said to depend on neural state; without neurological events, there would be no pain. The Gate Control Theory thus illustrates a weak form of ontological holism.

Conceptual Change

The ontological commitments of the Gate Control Theory are an intriguing example of the kind of conceptual change that comes about through scientific research. We come to understand our feelings of pain as complicated neural events. This is not as strange as it sounds at first. We understand hot and cold as the mean kinetic energy of molecules, and hot and cold are, in the first instance, feelings. There are not two things—heat and mean kinetic energy—there is but one, mean kinetic energy. The conceptual change instituted by theories like the kinetic theory of heat or the Gate Control Theory of pain are changes where two distinct concepts are unified.

To be continued…

Theory Critique by Parody?

The other day, I was looking at a model of transition to motherhood   class of nursing PhD students.  The authors had used grounded theory to identify stages in the process by which women entered the role of being a mother.  As we were puzzling over the diagram — a group of labeled, overlapping circles — one of the students remarked that “this is just like the transition to being a PhD student.”  We laughed, and indeed, all of the categories fit perfectly.  First you felt unready and lost, then you figured it out, and finally you were comfortable; transition complete. 

The categories of the model, it turned out, fit any transition whatsoever.  They captured “transition to pizza delivery drver” just as well as “transition to motherhood.”  The problem is that the categories were vague, even metaphorical, and that the relationships among the stages was not articulated.  One stage was said to “cause” another, but there was no description of what this cause amounted to.

So, here’s the lesson: if there is an absurd application of your model, something is wrong.  If your model can be parodied by making it easily fit pizza delivery or window washing, when it is supposed to be about a specific social process, then model is under-specified.  A model that applies indiscriminately says nothing.

 

People as Measurement Devices

Measurement and qualitative research

In his presentation at the 2009 International Philosophy of Nursing Conference in Bristol, John Paley argued that qualitative research could be understood as a kind of measurement. I was reminded of this provocative thesis when I read a new essay by Gualtiero Piccinini “First-Person Data, Publicity & Self-Measurement” (Philosophers’ Imprint, Vol 9, 2009). Piccinini engages the literature that tries to defend a “science of consciousness.” He criticizes the idea that first-person reports create a special kind of science, but concludes by accepting first-person data (when rightly understood) as scientifically legitimate and valuable.

These lines of argument challenge two common nursing conceptions of qualitative research. First, they challenge the characterizations of qualitative research that use measurement as a defining feature of quantitative research. Second, they challenge the common idea that the qualitative investigator is the measuring instrument.

First-person data and measurement

Paley’s argument – and note, this is based on my memory of his presentation in Bristol, so it may not accurately represent his view – assimilated qualitative research to a larger conception of measurement. Measurement is a form of representation, and the use of numbers to scale the measurement is a special case. Paley’s presentation described some interesting techniques for measuring attitude and belief that are semantically sensitive in the sense that they locate the subjects’ belief (attitude) in a semantic network. There is no numeric scale, yet there is an apparently reliable representational scale.

Piccinini is concerned with more with the scientific value of direct, first-person reports. He is arguing against those who take first-person reports of, say, pain or mental imagery to be private, yet legitimate scientific data. Those who support such a view hold that because first-person reports are private, any science which relies on them must have a different character than the familiar sciences with their public, reproducible data.

Puccinini challenges the notion of private scientific data on several grounds, and he ultimately rejects the idea that “the validity of first-person data is untestable by public means” (p. 10, cf. p. 3). His positive view is that first-person data are legitimate scientific data because they are public. Pain reports are a kind of first-person data that is essential to nursing research. One of the mistakes behind supposing that such reports are private is to suppose that the person who is in pain is an observer of the pain. Puccinini suggests that the subjects who are in pain are not the observers: the investigators are. The investigator measures (and thereby observes) the pain through the first-person reports of the subject. Thus:

A subject generating first-person behaviors to fulfill the purposes of a scientific observer is a self-measuring instrument. When a subject generates first-person behaviors, she embodies not only (part of) the experimental materials but also (part of) the measuring apparatus. (p. 11)

A person pointing at a degree of a pain scale thus provides an observation of pain in much the same way as a Geiger counter provides an observation of radiation.

Consequences for qualitative research

Paley and Puccinini are exploring an interesting and new way of assimilating qualitative and quantitative research. If they are right, then “measurement” does not discriminate between qualitative and quantitative research. Interviews (even if conceptualized as phenomenological researchers do) are measurements in much the same sense as glucose assays.

Even more interesting is Puccinini’s argument that the scientist is the observer, not the subject who reports the pain (etc.). Qualitative researchers have often said that the investigator is the measuring instrument of qualitative research. This is supposed to mean that investigator is to record her own responses to the interview, along with what is said (or done). The interpreter’s impressions are the data for the qualitative research. This is a mistake, if Puccinini is right. The researcher is not studying herself, she is studying a population (albeit a small one). They are the ones with the experiences, and their first-person reports of those experiences are the basis of any qualitative research. Puccinini’s essay is a nice reminder that qualitative research is more than autobiography.