Content uploaded by Sandra B Dunbar
Author content
All content in this area was uploaded by Sandra B Dunbar on Oct 14, 2014
Content may be subject to copyright.
This is the penultimate draft of the essay published in Philosophy of the Social Sciences, 31 (2001): 40-59.
Please cite the published version.
Methodological Triangulation in Nursing Research
Mark Risjord
Margaret Moloney
Sandra Dunbar
1. Introduction: Nursing Research and the Philosophy of Science
Comforting a patient at the bedside, taking a blood pressure in the examining room, or
instructing a new mother about breastfeeding, nurses are omnipresent in contemporary health care.
Few realize that this frontline work is supported by an autonomous research discipline. Nursing
research addresses theoretical and practical problems important to clinical nursing practice. The
research domain thus encompasses the full range of human responses to health and illness, including
biological and physiological manifestations as well as the emotional, motivational, psychological and
social facets of illness. The breadth of these questions motivates the use of a wide range of scientific
methods. Because nursing exists within a health care tradition heavily imbued with biomedical
theorizing, the methods of epidemiology and the biological sciences have been important tools.
Nursing is also interested in human response to illness — what the illness means to the patient —
and this motivates the use of interpretive approaches akin to ethnography. Sociological methods of
survey and interview are also widely deployed. Nursing thus overlaps methodologically with a
number of other disciplines.
The use of methods drawn from both the natural and the social sciences raises a number of
philosophically interesting issues. Some of these are being debated by nursing researchers
themselves, and this essay will take its point of departure from one such debate. Nurse researchers
confront phenomena that admit of, and perhaps require, the application of several different
methodologies. The study of patient responses to stress, for instance, may proceed through
2
physiological assays (analysis of saliva samples, for instance) or through in-depth interviews with the
patients. “Methodological triangulation” is the use of two or more different kinds of methods in a
single line of inquiry. Nurse researchers have asked about the relationship between these different
methods. Must different kinds of method support different kinds of theory, and thus produce two
distinct views of the phenomenon? If so, then a study based on “triangulation” is nothing more
than two studies packaged as one. Alternatively, can different methods be integrated into a single,
unified inquiry? If so, then methodological triangulation might yield results that could not be
obtained by either method alone.
There is a clear parallel between these questions and one of the central issues in the
philosophy of the social sciences: the relationship between the “natural” and the “social” sciences.
One side of this longstanding discussion is that the natural and social sciences are methodologically
distinct. This view has been called “methodological separatism” (Henderson 1993). For a variety of
reasons, separatists have held that human beings and human societies require methods of inquiry
that are different from those methods used to study the non-human world (Collingwood 1946;
Davidson 1984; Dilthey 1996; Taylor 1971; Winch 1958). The issue is not about the legitimacy of
the social sciences — a separatist holds that methods of the human sciences are as epistemically
legitimate as those of the natural sciences. The question is whether there is a natural divide among
kinds of methods and the theories they support.
In the debate over methodological triangulation, some nurse researchers have adopted a
separatist position concerning the use of different methods in nursing research. On a separatist
view, interviews with patients under stress produce a very different kind of theory than, say, an
epidemiological analysis of their saliva samples. At best, the two are complementary theories of a
complex phenomenon. Others have rejected separatism, holding that widely different methods can
be productively combined in a unified study of a phenomenon like stress or pain. In this essay, we
3
will begin with a close look at the debate over methodological triangulation as it arises in nursing.
Then, we will make explicit some of the philosophical presuppositions that underlie the debate.
Parties to this debate share a view about the structure of an empirical theory, and corresponding
presumptions about how theories are confirmed by evidence and how they are related to
background assumptions. These views have been subject to criticism in the philosophy of science,
and we will bring the results of this criticism to bear on the nursing debate. We will argue that if
some underlying assumptions are modified, the debates about triangulation can be resolved. This
conclusion is not only of interest to nursing researchers. In the concluding section, we will suggest
that it has ramifications for the way that philosophers think about how the “natural” and “social”
sciences are related.
2. The Methodological Triangulation Debate
Nurse researchers have traditionally divided their methods into two kinds: quantitative and
qualitative. Roughly, qualitative methods are interpretive and textual. Examples of qualitative
methods would include passive observation of the way in which patients and their families interact,
participant observation of nurses and patients in a particular hospice, open-ended interviews with
patients, or the analysis of patient diaries. Quantitative methods are drawn from the natural and
biological sciences. They include statistical analysis of procedure outcomes, biological markers, or
questionnaires. Data for a typical quantitative method is expressed numerically and is collected by
some means of measurement. The debate over methodological triangulation in nursing has focused
on the relationship between “qualitative” and “quantitative” methods.
i
Before jumping into the
debate, it will be helpful to have some examples of the ways in which nurses have triangulated with
these different kinds of method.
Qualitative research methods are often used to explore a phenomenon that has not been
previously well described. The results are then used to develop survey instruments. The survey
4
instruments yield results that can be analyzed statistically. Fleury’s research provides an example
(1991; 1993; 1998). Fleury focuses on the concept of self-regulation in relation to cardiac risk
reduction. Self-regulation is a process of controlling one’s motivations through selectively
processing information, monitoring one’s own behavior, judging performance, and self evaluation.
Self-regulatory mechanisms are thought to be crucial to understanding the volitional aspects of
behavioral change. They reflect the ways in which people attempt to behave in accordance with
personal health goals. Fleury (1991) initially used qualitative methods and interviewed 29 individuals
who were attempting to initiate and sustain programs of cardiac risk factor modification. These
qualitative data were analyzed into six dimensions in which individuals regulated self and
environment. These six dimensions and the qualitative data were then used to frame items
developed in a 6-point Likert scale for a quantitative instrument, the Index of Self Regulation (Fleury
1993; 1998).
A different sort of example where patient interviews were integrated with quantitative
measures is a recently completed pilot study on migraines in perimenopausal women (Moloney MS).
It is currently believed that hormonal migraines are due to estrogen shifts during this transitional
time, but very little work has addressed this problem. There is also research showing that quality of
life in migraineurs in general is as poor as that of persons with other chronic diseases such as
diabetes mellitus. A qualitative approach was employed that used interviews and a questionnaire
with open-ended questions. This was supplemented with a questionnaire where the answers were
standardized by numeric scales, and a standardized quality-of-life questionnaire. The patients also
kept a diary that combined both quantitative and qualitative items. In this pilot study, themes in the
qualitative data centered around keeping on with one’s life in spite of headaches, and refusing to
permit headaches to interfere with daily life. However, the standardized quality-of-life
5
questionnaires revealed unexpected anger and resentment about having to cope with headaches.
This information, not apparent in the interviews, helped to clarify and deepen the qualitative data.
Sometimes, the project is driven by an explicit, explanatory theory and complemented by
qualitative methods. Dunbar et al. (1999) provides an example of this form of triangulation. This
study examined factors associated with the patients’ recovery process within the first three months
after they had received an implantable cardioverter defibrillator (ICD). These patients were
considered at risk for cardiac arrest from abnormal heart rhythms, and approximately 25% had been
successfully resuscitated from cardiac arrest while the remainder were considered at high risk for a
life-threatening arrhythmia. The internal defibrillator, a box that was surgically implanted beneath
the skin of the chest wall like a pacemaker, prevented cardiac arrest by monitoring the heart rhythm
and delivering a shock to convert the heart back into a normal rhythm. A variety of standardized
questionnaires were used to evaluate the patients’ adjustment to this procedure. Qualitative
interview data were also coded and compared with scores obtained from a questionnaire that
measured mood states. One of the unexpected findings was that patients who had experienced
sudden cardiac arrest had a better mood level one month after the procedure than the group that
had not. This finding was explained by the qualitative data, which showed that the ICD provided a
sense of security for this group, whereas those who had not experienced sudden cardiac arrest
expressed more uncertainty about the need for the device.
There are three rationales frequently given for using methodological triangulation. The first
is completeness. Quantitative methods can further develop findings derived from qualitative research
(and vice versa). The methods complement each other, providing richness or detail that would be
unavailable from one method alone. This is exemplified by both the Dunbar and Moloney studies
above. The second might be called abductive inspiration. As in Fleury’s research, described above,
qualitative research is often used where a phenomenon is poorly understood. Interviews with
6
patients can orient the investigators to the material. The results may suggest hypotheses to be tested
by quantitative methods. Qualitative investigation can also help organize quantitative data that has
already been gathered or suggest ways new of approaching the phenomenon. The final, and most
controversial, rationale for triangulation is confirmation. In its most modest form, qualitative methods
can clarify the results of quantitative research, such as apparently inconsistent findings. More
tendentiously, qualitative and quantitative results are sometimes thought to support each other.
Triangulation would thus yield a stronger result than either method could yield alone.
The debate about triangulation on which we will be focusing has occurred among those
nurse researchers who think that triangulation is possible and useful. The issues concern the
underlying rationale for mixing methods and the degree to which the methods can be “mixed” at all.
This position, which we will call the “blending view,” contends that methodological triangulation
can yield all three benefits mentioned above: completeness, abductive inspiration, and confirmation.
Critics respond by arguing that qualitative and quantitative methods are based on deeply different
assumptions about the objects studied. Proponents of this “building block view” thus contend that
different methods and their results must remain independent. Since the methods do not support
each other, triangulation can not yield confirmation. Triangulation is useful, but only for
completeness and abductive inspiration. The principle point of contention between the building
block view and the blending view is thus whether the use of different kinds of methods can together
confirm the results of a study to a greater degree than either method alone.
Most parties to the debate over methodological triangulation take methods to be closely tied
to the kinds of theories they support, as well as background assumptions and standards for
evaluation. The result is similar to Kuhn’s conception of a paradigm, and nurse researchers have
adopted this language. Qualitative and quantitative methods are said to be embedded in different
paradigms (cf. Haase and Myers 1988). The quantitative and qualitative paradigms have distinct
7
motivations and goals of inquiry. In articulating their blending view, Haase and Meyers align the
quantitative—qualitative distinction with the distinction between explanation and description:
In order to gain understanding, the empirical-analytic paradigm has emphasized
confirmation of theory by explaining. The human science paradigm, also in search of
understanding, has emphasized discovery and meaning by describing. (Haase and Myers
1988: 130)
Part of what is being expressed by the use of the word “paradigm” is a difference in the structure of
theory. Theories supported by quantitative data are hierarchically structured systems of laws. The
explanation of a phenomenon or a regularity is a deduction of that phenomenon from laws, initial
conditions, and bridge laws. It is thus no surprise to find nurse researchers referring to the paradigm
that drives quantitative methods as “positivist” (Duffy 1987; Haase and Myers 1988; Shih 1998).
Both the building block theorists and the blending theorists argue that the quantitative and
qualitative paradigms include fundamental ontological presuppositions about the object of inquiry,
and the relationship between the inquiring subject and her object. Quantitative inquiry presupposes
that its object and all of the relevant properties and relationships can be measured. Qualitative
inquiry takes its object to be a subject — a being for whom things can be meaningful and valuable.
Finally, the qualitative and quantitative paradigms encompass different evaluative orientations.
Methodologically, each has its own global criteria according to which any scientific work is to be
evaluated, and specific criteria for the evaluation of particular kinds of research. Quantitative
research is evaluated in terms of concepts like predictive power and statistical significance. Since
these notions do not apply comfortably to interpretations, qualitative research is evaluated in terms
of its internal coherence and the degree to which the subjects think the result is an accurate portrayal
of their experience.
Building block theorists argue that the depth and the nature of the differences between
qualitative and quantitative paradigms make them incommensurable. First, they argue that there is a
8
difference in the way that theory and data are related within qualitative and quantitative paradigms.
Morse, a building-block theorist, writes:
Thus, in methodological triangulation, the key issue is whether the theory that drives the
research … is developed inductively from the research per se or used deductively as in
quantitative inquiry. (Morse 1991: 121)
What Morse means by “inductive” and “deductive” is clear enough from the context.
ii
Deductive
research begins with a theory, ideally articulated as a body of laws or universal generalizations. A
hypothesis is derived from this theory, and it is tested against observation. Inductive research begins
with observations alone — untheorized and without hypotheses. It proceeds to distill descriptive
claims from the observations. Theorists like Morse take this difference to entail a difference in the
kinds of theory that can be supported by qualitative and quantitative methods. Quantitative
methods are necessary to support explanatory theories that are structured as hierarchies of laws.
Qualitative interpretations are variously said to be “holistic,” “dynamic,” or “subjective” (Dootson
1995; Duffy 1987; Haase and Myers 1988). Interpretations, which are supposed to be neither
explanatory nor nomological, must be supported by different kinds of methods. Finally, the
building block theorists argue that the standards of evaluation in each paradigm yield nonsense when
applied to the other. Interpretations are not appropriately evaluated in terms of statistical
significance, and it makes no sense to ask whether the analysis of a blood sample coheres with the
blood’s point of view.
The building block theorists take these differences in the relation of theory to data, the
structure of theory, and the standards of evaluation to entail that the qualitative and quantitative
components of a triangulation study must be kept distinct. Failure to do so will result in
incoherence or outright inconsistency (Dootson 1995; Foster 1997; Morse 1991). In a triangulation
study, the qualitative and quantitative components must support theories with different logical
structures. Moreover, an individual research method should be expected to withstand scrutiny on its
9
own merits. These proponents of triangulation thus insist that the methods be kept distinct and the
results of each judged on their own criteria. It follows that confirmation cannot be an appropriate
goal of triangulation. There is no hypothesis or set of hypotheses that could be supported by both
methods. Triangulation can be used only for abductive inspiration or for completeness.
Blending theorists agree that there are deep differences between qualitative and quantitative
research, but they deny that these differences stand in the way of confirmation. Knafl (1991; 1988),
Haase (1988), Duffy (1987), Mitchell (1986) and others hold that methodological triangulation
increases the reliability, validity, or accuracy of the study. There seem to be two primary arguments
in the literature. First, they argue that methods have different strengths and weaknesses (Duffy
1987: 132; Mitchell 1986: 22). Qualitative research tends to focus on a small group of participants.
For this reason, it is difficult, sometimes impossible, to generalize the conclusions. Qualitative
results can only be applied to new individuals when the context is taken into account. Triangulation
with a quantitative methodology can increase the researcher’s confidence that her conclusions apply
to a larger population. Quantitative methods have their own limitations. The emphasis on
standardization and generalizability means that the questions are often narrow. A survey may not
capture what is significant to the patient. Consequently, numeric results can be superficial or
misleading. Correlation with qualitative interviews can increase the researcher’s confidence that a
survey has uncovered a meaningful result. Quantitative and qualitative methods each compensate
for the other’s weaknesses, and a therefore increase the researcher’s confidence in the whole study.
A second argument deployed by blending theorists is that triangulation helps eliminate bias
(Duffy 1987: 132; Mitchell 1986: 21). Both qualitative and quantitative research is subject to bias.
Since the biases occur in slightly different ways, the use of both methods enhances the researcher’s
confidence that the biases have been identified. For instance, in both kinds of research, the subject
may answer in socially desirable ways or in the way she thinks the researcher wants the question to
10
be answered. In this respect, the important difference between qualitative and quantitative research
is in the relative distance between subject and researcher. In a face-to-face interview, the researcher
may inadvertently express her attitude through tone, posture, or facial expression. At the same time,
the researcher is able to read the body language of her interlocutor, and thus identify questions that
make the subject uncomfortable or evasive. A pencil-and-paper questionnaire is relatively
anonymous, so, the researcher’s presence does not influence the answers. Correspondingly, it is
difficult for a quantitative method to determine whether the answers were authentic. Quantitative
and qualitative research are thus subject to the same bias in complementary ways. Where the
quantitative and qualitative results diverge, the results of one can be used to refine the questions (or
the questioner’s technique) of the other. If the results of the methods converge, then the researcher
can be more confident that social desirability bias is not present. Triangulation therefore increases
confidence in the results to a higher degree than the use of either method alone.
3. Analysis of the Debate
It is interesting to note that the blending theorists’ response to the arguments against
confirmation do not challenge any of the fundamental presuppositions of the argument. Rather,
they point out ways in which the use of multiple methods might enhance confirmation in spite of the
deep differences between what they take to be paradigms of research. We will argue that this is a
tactical mistake on the part of the blending theorists. What really needs scrutiny is the idea that the
differences among research methods amount to different paradigms. This proposition depends on
at least four different presuppositions. When made explicit, it is clear that all are dubious. Our
strategy in the remainder of this essay will be to argue that different commitments entail a more
satisfactory conception of methodological triangulation. The result will be a version of the blending
view that is more robust than previous versions precisely because it does not buy into key claims
about differences among “paradigms.”
11
As is obvious from the foregoing discussion, nursing researchers presuppose that:
1. Qualitative and quantitative methods are distinct.
Considering the matter historically, it is true that methods called “quantitative” and “qualitative”
have arisen from different traditions. Quantitative research is inspired by the natural sciences and
those parts of the social sciences that have empiricist leanings. Qualitative research in nursing was
developed partly in conscious opposition to such empirical methods. Drawing on the hermeneutic
and phenomenological traditions in philosophy, qualitative researchers argued that the methods of
the natural sciences could not elucidate realms of experience crucial to the enterprise of nursing
research. The experience of the patient, the meaning of a treatment, and the value of health and
illness to the patient must elude detection by questionnaires or biological assays. Qualitative
research develops interpretations using means that would be familiar to a literary critic,
ethnographer, or historian. In practice, however, the boundaries between qualitative and
quantitative research are fuzzy. Interviews are often coded, and the analysis of patterns of word
usage is used to support the interpretation. Does coding an interview or videotape make it
“quantitative”? Or is it still “qualitative”? Questionnaires are textual, and the words notoriously
require careful interpretation, both by the subjects and the researcher. Yet, they yield numeric
results that are treated with statistical tools. To add to the confusion, surveys based on
questionnaires are sometimes considered the qualitative component of a mixed methodology study.
In spite of such vagueness in practice, nurse researchers take the distinction to be
conceptually clear and deep. It is hard to deny that there are substantial differences among research
methods. One might wonder whether the concepts “qualitative” and “quantitative” capture these
differences in an interesting way, but we will not pursue that line of inquiry. The mere distinction
between qualitative and quantitative methods is only a small part of what nurse researchers mean
12
when they talk of qualitative and quantitative paradigms. The full force of this notion is provided by
three additional presuppositions.
Nurse researchers extend the qualitative-quantitative distinction from methods to theories.
They thus presuppose that:
2. Qualitative and quantitative methods support different kinds of theory.
This presupposition is required by the building block view. The “blocks” are bodies of theory, and
on this view they must remain distinct. There must be, then, some sort of difference — either in the
kind of content or the logical structure of the theory — between the quantitative and the qualitative
components. Typical examples of theories called “quantitative” are nursing theories that draw
heavily on the biological sciences, epidemiology, or sociology. Typical examples of theories (or
better, interpretations) that are called “qualitative” bear a strong resemblance to ethnography or
history.
Quantitative theories are conceived in traditional empiricist terms as a body of universal
generalizations. Two aspects of the empiricist model of science constitute crucial presuppositions of
the triangulation debate:
3. Quantitative theories are structured as hierarchies of deductively related
propositions.
4. Quantitative methods confirm hypotheses one-by-one.
These are less obvious elements of the nursing debate than the first two presuppositions.
Nonetheless, they surface explicitly when nurse researchers are discussing quantitative theory and
method. For example, when contrasting qualitative and quantitative research paradigms, Duffy
writes:
The procedures employed by the quantitative researcher are usually highly structured and
designed to verify or disprove predetermined hypotheses. (Duffy 1987: 131)
13
According to Haasse and Meyers, quantitative theories are distinguished by their use of law-like
generalizations that have predictive value (Haase and Myers 1988: 135). This conception of
quantitative theory is reflected in the nursing research methodology textbooks. In Chin and
Kramer’s widely used text, they write:
[Research to validate a theory] is usually thought of as a deductive approach. The research
starts with an abstract relational statement derived from theory. From the theoretic
statement, hypotheses or research questions are created for a specific research situation.
(Chin and Kramer 1999: 125)
Chin and Kramer also write that:
A single study is usually based on one or two relational statements from among several that
might possibly be extracted from a theory. No one study can test the entirety of a theory.
(Chin and Kramer 1999: 124)
Clearly, then, the model of a quantitative theory is one in which a theory is a set of generalizations
linked by deductive relationships. More general propositions entail less general propositions. A
theoretical proposition is tested by deriving observational statements from it. Theoretical
propositions are rejected when their observational consequences are false.
This model of theory structure is quite familiar from mid-twentieth century philosophy of
science. As expressed by presuppositions (3) and (4), the model is a crucial part of the notion that
there are deep differences between qualitative and quantitative paradigms. Nurse researchers have
held that qualitative methods cannot confirm quantitative claims, and central examples of
quantitative claims would be the correlation of two values or a causal relationship between two
variables. Assuming (3) and (4), confirmation of a quantitative theory must proceed by deducing
values for these variables, given initial conditions. Propositions of a quantitative theory thus cannot
be tested without some kind of measurement device, and nurse researchers have taken this to be the
mark of a quantitative method. Quantitative theory thus requires confirmation by quantitative
methods.
14
It follows from these four presuppositions that there is a very close relationship between a
method and the kind of theory it can support. This idea has profoundly shaped the debate over
methodological triangulation. Both sides agree that qualitative methods could not confirm a
quantitative theory, and vice versa. The issue about confirmation, then, is whether and how
quantitative and qualitative theories of a single phenomenon might support each other. Blending
theorists point to complementary strengths and resistance to bias. However, in the light of the
foregoing analysis, it is no longer clear how these could contribute to confirmation. The kinds of
theory are logically independent in the sense that the propositions of the quantitative theory do not
entail (nor are entailed by) propositions of the qualitative theory. Therefore, given presupposition
(3), there is no larger theory, encompassing both the qualitative and quantitative components, that
could be confirmed or disconfirmed. By consenting to the terms of the debate, blending theorists
have made it difficult to defend their idea that methodological triangulation might yield some kind
of confirmation.
4. Confirmation and the Structure of Theory
The result of the foregoing section shows that, if triangulation is to yield confirmation, then
some of the presuppositions of the debate need to be challenged. Presuppositions (3) and (4) are
likely targets, both because they have been subject to so much criticism, and because they have well
developed and plausible alternatives. One of the alternatives to (3) is a coherence conception of theory
structure. The coherence conception differs from the classical, hierarchical view in several ways.
The classical view is foundationalist in the sense that observation statements constitute the ultimate
justifiers of scientific theory. Predictions are observation statements that are derived from theory.
Where the prediction fails, some of the theoretical claims used in its derivation must be rejected.
Theoretical statements receive inductive support from some body of observation. If those
observations are found suspect, then the theoretical claims which they support must be suspect too.
15
A theoretical statement, then, is scientifically justified if it is supported in the right way by
observation statements. On a coherence view, a statement is justified if it coheres with the other
statements of the theory, including observation statements. A coherence theory is thus not
hierarchical because justificatory support does not run from the observations to the most abstract
laws. Rather, the theoretical statements and the observation statements hang together in
relationships of mutual support.
The classical view of theories took theoretical claims to be connected by inferential relations.
A theory was ideally conceived as a partially axiomatized formal system. On the version of
coherence theory we are proposing here, statements of a theory are related as question and answer.
For any proposition of the theory, P, the inquirer can ask “Why P?” If the answer to this question is
Q, then the theoretical statements P and Q will cohere. Risjord (2000) took why-questions to be the
primary kind of question that relates propositions within a coherent theory. Others, such as
Thagard (1992; 1999) and Sintonen (1989) are more pluralistic with respect to the forms of question
that are taken to structure theory. Differences among these views are not relevant to this essay. If
we identify an explanation with the raising and answering of a question, as many authors do
(Garfinkel 1981; Lipton 1991; Risjord 2000; van Fraassen 1980), then the coherence in question is
explanatory coherence. In a coherent theory, each theoretical statement is either the explanation of,
or gets explained by, some other theoretical statement.
Finally, as expressed by presupposition (4), the classical conception of theory structure took
confirmation to be a relationship between an individual hypothesis and a definite domain of
observation. Low-level hypotheses are directly confirmed or disconfirmed by observations, and
more abstract theoretical statements are confirmed or disconfirmed insofar as they entail these low-
level hypotheses. On an explanatory coherence view, theories are confirmed or disconfirmed as
wholes. If a low-level hypothesis is rejected, other hypotheses that previously depended upon it may
16
be preserved, if they continue to cohere with the rest of the theory. The question of how a
hypothesis is confirmed is replaced by the question of how theories are chosen: Under what
conditions should one theory be preferred to another? The answer, roughly, is that the better theory
is more coherent. One theory is more coherent than another (again, roughly) if it leaves fewer
questions unanswered. This means that the better theory will be more complete — it will account
for more of the observations.
For the purposes of this essay, there is a crucial upshot of the coherence understanding of
theoretical structure. A coherence view does not demand that every theoretical statement be a
natural law, nor every question be a causal question. The question-and-answer theoretical structure
can be found in the interpretation of meaning, value, norms, and experience (Risjord 2000).
Questions about meaning or lived experience do not arise concerning quantitative data, nor do
causes answer questions about human values or norms. Hence, theories about these sorts of
phenomena make different claims than theories concerning, say, biological responses to stress. In
spite of the different questions asked and claims made, there is no logical reason why theories about
these different domains may not be combined into a single theory. On an explanatory coherence
view, the structure of all theorizing is question-and-answer. A biological theory and a humanistic
interpretation of stress, for example, would be combined into a single theory if questions about the
biology could be answered by claims about experience of the patients and vice versa. The whole
would be more coherent than either part alone, so integrating the two would increase the coherence
of the whole theory.
Rejecting presuppositions (3) and (4) and adopting a coherence account of theory structure
and confirmation has important consequences for the methodological triangulation debate. Since all
theory is structured as a body of claims made coherent by relationships of question and answer,
there are not deep, logical differences between qualitative and quantitative theory. This is not to say
17
that there are no differences between the theories that provide paradigmatic examples of
“quantitative” and “qualitative” theory, only that these are not logically different kinds of theorizing.
It follows that presupposition (2) — that qualitative and quantitative methods support different
kinds of theory — must be rejected. The absence of presuppositions (2), (3), and (4) breaks the link
between theory and method that had been assumed by all parties to the methodological triangulation
debate. A given theory does not require a particular kind of method. There are, of course,
differences among methods.
iii
And the best answer to a given question must be determined with
methods appropriate to the question. The choice of qualitative or quantitative methods, however, is
no longer a global choice in the sense that it determines (or is determined by) the researcher’s whole
“paradigm.” Rather, it is a local choice that depends on the particular question an inquirer is asking.
A theory may be composed of a number of questions for which different kinds of methods are
suitable. Where this is the case, methodological triangulation will be necessary.
5. How to Blend Different Research Methodologies
Methodological triangulation can thus be conceived as the use of different kinds of methods
to answer different questions arising from a single theory. When we look at examples of
triangulation in nursing research, we would expect to find the results of qualitative and quantitative
research methods building on each other. Questions answered by qualitative methods will have to
arise out of answers generated by quantitative methods, and questions demanding quantitative
methods will have to arise out of answers generated by qualitative methods. Looking back at the
examples from section 2, this is precisely what we find.
Qualitative nursing research builds on the results of quantitative methods in obvious ways.
Nursing research occurs in a health care context where the questions are typically causal and
answered by quantitative methods. This quantitative research provides the topics and
presuppositions of qualitative questions. In Moloney’s migraine study (Moloney MS), she
18
administered a standardized quality-of-life questionnaire. This instrument yielded a numeric score
for things like anger and resentment. The point of this questionnaire was to understand how
migraineurs compared to persons who suffered from other conditions. This quantitative method
yielded a surprisingly high level of anger and resentment. It prompted Moloney to ask why
migraineurs were feeling this way, and how these feelings motivated the migraineurs in their ways of
coping with the migraines. Answers to these questions drew on interviews and open-ended
questions in the patient’s daily diary. The questions prompted by quantitative data received answers
provided by qualitative methods.
On a building block model of triangulation, such as Morse’s (1991), Moloney’s use of
quantitative methods is understood as providing only a bit of detail to a qualitative study. The
understanding of triangulation we are presenting exposes the deeper connections among the
methods. Moloney’s questions about the patient’s anger may not have arisen in the absence of the
quantitative instrument. The quantitative research thus provided a basis for qualitative questions.
Answering these questions by interviews and diary analysis confirms the results of the quality of life
questionnaire by making these results cohere with other things known about migraineurs. Here we
have an understanding of triangulation that lives up to its name — the two methods provide a
perspective unattainable by using one alone.
The results of qualitative research can similarly provide the topics and presuppositions of
questions that need numeric answers. Nurse researchers often ask questions that inquire into the
patient’s point of view. They ask what the experience of a particular kind of patient is like, or how
families respond to the illness of a member. These questions are naturally answered with interviews
or other textual data gathering methods. One result of these questions is that the nurse researcher
becomes aware of the ways in which the patient is conceptualizing her illness. Any question has to
mobilize some concepts. Qualitative research provides concepts that are drawn from the patient’s
19
experience. Thus, by deploying these conceptions, questions that demand quantitative answers may
be more precise, marking distinctions that are relevant to the patient’s motivation, emotional
outlook, or behavioral response. The researcher’s confidence in the numeric scores is increased by
the fact that the concepts appearing in the questions cohere with other aspects of the theory. The
studies by Fleury (1991; 1993; 1998) and Dunbar (1999) reveal this sort of interplay among
questions. As discussed above, Dunbar’s research concerned patients who had received an implant
that would shock the heart back into a normal rhythm should an abnormal rhythm begin. A
standardized instrument for measuring mood revealed, unexpectedly, that patients who had
experienced sudden cardiac arrest had a better mood level than those who had not. This
quantitative result raised questions that were naturally answered by qualitative means. Interviews
with patients revealed that the presence of the implant gave them a sense of security. Security is a
concept that might not have arisen except for the interviews. Once in place, it may become the
topic of questions that ask for quantitative answers. For instance, what is the relationship among
cardiac patients’ sense of security, illness appraisal, other treatment outcomes, and severity of
arrhythmia?
In Section 2, we argued that confirmation was the pivotal issue between the building block
and the blending views of triangulation. Building block theorists insisted that triangulation can only
be used to increase the completeness of a study, never for enhancing confirmation. Blending
theorists held that triangulating methods resulted in mutual confirmation. Given their
presuppositions, both sorts of views had difficulty understanding how the results of different
methods could hang together in a single theory. Adopting a coherence understanding of theory
structure and confirmation permits the resolution of these issues. Since confirmation is a matter of
maximizing coherence, and since a more complete theory is more coherent, the goals of
confirmation and completeness collapse into one another. On a coherence view, the building block
20
theorists’ goal of comprehensiveness amounts to mutual confirmation. Both the quantitative and
qualitative blocks are parts of a theory that is better confirmed than either of its parts (because the
whole is more complete). The two kinds of research hang together into a single theory because
quantitative and qualitative methods are now understood as different ways of answering questions.
Insofar as these questions and their answers interlace into a coherent body of theory, both kinds of
method can contribute. A coherence understanding of theory structure thus supports a blending
view of triangulation that goes much farther toward revealing the real value of triangulation than
other blending views.
6. Conclusion: the Fate of Separatism in the Philosophy of Social Science
What lessons can a philosopher of social science take home from the foregoing discussion of
triangulation in nursing theory? The dispute over methodological triangulation both mirrors and
draws upon the philosophical debate about the relationship between the natural and social sciences.
The grounds for holding that the natural and social sciences must have different methods are
diverse. The issue about the structure of theory, on which we have focused, is only one facet of the
issue. Others include the character of explanation, the role of causality in understanding human
behavior, and the relevance of norms and meaning to scientific inquiry. The philosophical discussion
of these issues has typically contrasted different disciplines, such as physics and history. Separatism
is challenged by nursing research, because the domain of inquiry seems to need both natural
scientific and interpretive approaches. The first ramification of attending to nursing research is that
the debate over the relationship of the natural and social sciences needs to respond to domains
where there are live issues about how different sorts of methods might fit together.
A second, deeper, ramification of the forgoing discussion is that separatism has
consequences for the way in which nursing research is conducted. If the picture of nursing inquiry
resulting from separatism is unsatisfactory, and if better views of nursing inquiry are made possible
21
by alternatives to separatism, then separatism itself has been found wanting. The proponents of the
building block view of triangulation have explicitly adopted a separatist understanding of the
relationship between the natural and social sciences. Moreover, the debate among nursing theorists
over triangulation has presupposed many elements of a separatist view. A consequence of
separatism is that quantitative and qualitative methods could not mutually support one another.
While this is not a dire consequence, it is an unfortunate one. It is independently plausible that
convergence among results of different methods should make a study more reliable. Wimsatt, for
instance, has called this phenomenon “robustness.” In his essay, “Robustness, Reliability, and
Overdetermination” (1981), Wimsatt argued that the multiple determination of results is found in a
variety of scientific contexts. Throughout scientific inquiry, robustness provides grounds for
judgments of the objectivity of postulated entities, and it permits confidence in heuristics that are
known to be fallible. The possibility of determining results by multiple methods thus is desirable,
and a philosophical view of nursing research that precluded it is less plausible than a view that
permitted triangulation and provided its philosophical foundation.
It is our position that nursing inherited a mistake from the philosophy of social science. As
we argued above, replacing the hierarchical conception of the structure of theory with one based on
explanatory coherence undermines a crucial presupposition of the nursing debate. The result is a
blending understanding of triangulation that accounts for the way in which multiple determination
of a result can enhance confirmation. It also provides for a richer understanding of the way in
which nurse researchers have used qualitative and quantitative research, and gives guidance on the
practical issues that arise in these contexts. The upshot for philosophers of social science, then, is
that we need to radically reconsider the notion that the natural and social sciences have
irreconcilable methods.
iv
22
References
Chin, Peggy L., and Maeona K. Kramer. 1999. Theory and Nursing: Integrated Knowledge Development.
Fifth ed. St. Louis: Mosby.
Collingwood, R. G. 1946. The Idea of History. Oxford: Oxford University Press.
Davidson, Donald. 1984. Inquiries in to Truth and Interpretation. Oxford: Clarendon Press.
Dilthey, Wilhelm. 1996. Hermeneutics and the study of history. Translated by Makkreel, Rudolf A. and
Rodi, Frithjof. Princeton: Princeton University Press.
Dootson, Sally. 1995. "An In-Depth Study of Triangulation." Journal of Advanced Nursing 22 (1):183-
187.
Duffy, Mary E. 1987. "Methodological Triangulation: A Vehicle for Merging Quantitative and
Qualitative Research Methods." Image: Journal of Nursing Scholarship 19:130-133.
Dunbar, S. B., L. K. Kimble, L. S. Jenkins, M. H. Hawthorne, W. N. Dudley, and M. Slemmons.
1999. "Factors Associated with Outcomes Three Months after Internal Cardioverter
Defibrillator." Heart and Lung 28:303-315.
Fleury, J. 1991. "The qualitative generation of wellness motivation theory." Nursing Research 40:286-
291.
Fleury, J. 1993. "Preserving qualitative meaning in instrument development." Journal of Nursing
Measurement 1:135-144.
Fleury, J. 1998. "The Index of Self Regulation: development and psychometric analysis." Journal of
Nursing Measurement 6:3-17.
Foster, Roxie L. 1997. "Addressing Epistemologic and Practical Issues in Multimethod Research: A
Procedure for Conceptual Triangulation." Advances in Nursing Science 20 (2):1-12.
Garfinkel, Alan. 1981. Forms of Explanation. New Haven: Yale University Press.
Haase, Joan E., and Sheila T. Myers. 1988. "Reconciling Paradigm Assumptions of Qualitative and
Quantitative Research." Western Journal of Nursing Research 10 (2):128-137.
Henderson, David K. 1993. Interpretation and Explanation in the Human Sciences. Albany, NY: State
University of New York Press.
Knafl, Kathleen A., and Bonnie J. Breitmayer. 1991. "Triangulation in Qualitative Research: Issues
of Conceptual Clarity and Purpose." In Qualitative Research: A Contemporary Dialogue, edited by
J. M. Morse. Newbury Park: Sage Publications.
Knafl, Kathleen A., Marian M. Pettengill, Mary E. Bevis, and Karin T. Kirchoff. 1988. "Blending
Qualitative and Quantitative Approaches to Instrument Development and Data Collection."
Journal of Professional Nursing 4:30-37.
Lipton, Peter. 1991. Inference to the Best Explanation. London: Routledge.
Mitchell, Ellen S. 1986. "Multiple Triangulation: A Methodology for Nursing Science." Advances in
Nursing Science 8 (3):18-26.
23
Moloney, Margret. MS. "An Exploratory Study of Perimenopausal Women with Migraines."
unpublished manuscript.
Morse, J.M. 1991. "Approaches to qualitative-quantitative methodological triangulation." Nursing
Research 40:120-123.
Risjord, Mark. 2000. Woodcutters and Witchcraft: Rationality and Interpretive Change in the Social Sciences.
Albany, NY: SUNY Press.
Shih, Fu-Jin. 1998. "Triangulation in Nursing Research: Issues of Conceptual Clarity and Purpose."
Journal of Advanced Nursing 28 (3):631-641.
Sintonen, Matti. 1989. "Explanation: In Search of the Rationale." In Scientific Explanation, edited by P.
Kitcher and W. Salmon. Minneapolis: University of Minnesota Press.
Taylor, Charles. 1971. "Interpretation and the Sciences of Man." Review of Metaphysics 25:3-34, 45-51.
Thagard, Paul. 1992. Conceptual Revolutions. Princeton: Princeton University Press.
Thagard, Paul. 1999. How Scientists Explain Disease: Princeton University Press.
van Fraassen, Bas C. 1980. The Scientific Image. Oxford: Oxford University Press.
Wimsatt, William C. 1981. "Robustness, Reliability, and Overdetermination." In Scientific Inquiry in the
Social Sciences, edited by M. Brewer and B. Collins. San Francisco: Jossey-Bass.
Winch, Peter. 1958. The Idea of a Social Science. London: Routledge and Kegan Paul.
24
Endnotes
i
The nursing debate thus presupposes that there is a clear and useful distinction between “qualitative” and
“quantitative” methods. We will acquiesce in the distinction for the moment, and scrutinize it in the next section.
ii
Nurse researchers typically use “inductive” and “deductive” in this way. It is consonant with nineteenth
century usage (cf. the OED entry for “induction”), but importantly different from its use in contemporary philosophy.
iii
We may once again simply avoid the question of whether “qualitative” and “quantitative” is the best way of
capturing these differences. We will continue to use this terminology, but the argument will not require a sharp
distinction among kinds of method.
iv
The authors would like to thank Paul Roth, Alison Wylie, William Wimsatt, and the other participants in the
2000 Philosophy of Social Science Roundtable for their very helpful discussion and criticism.
