Académique Documents
Professionnel Documents
Culture Documents
approaches
by Katrin Niglas
http://www.leeds.ac.uk/educol/documents/00001544.htm
katrin@tpu.ee
Tallinn Pedagogical University
Narva mnt 25, Tallinn, 10120, ESTONIA
Introduction
Last year in Lahti I gave a paper concerning the debates (sometimes called as
"paradigm wars") about differences and similarities between quantitative and
qualitative research. I also presented the results of my small-scale investigation,
which showed that there were studies, which combined qualitative and quantitative
approaches in different ways. In this paper I want to look further and address some
problems concerning the use and integration of multiple methods in a social
scientific study.
In a long run there are three different widely advocated positions towards the
possibility and usefulness to use quantitative and qualitative approaches in
complimentary, combined or mixed ways:
The advocates of the first position, which I would call strong paradigmatic view,
declare that only one of those approaches is good/appropriate/scientific enough for
the inquiry about the social life. They say that quantitative and qualitative research
methodologies are tightly bound to different mutually exclusive epistemological
positions. From here follows that there is no point even to talk about the possibility
of combining or mixing of those approaches. The proponents of this position are
sometimes called purists.
The advocates of the second position, which I would call week paradigmatic
view, are somewhat more tolerant towards different methodologies saying that both
of them can be used and are useful, but as they carry with them different
philosophical underpinnings they are suitable in very different situations and
contexts and therefore one can not and should not mix or combine quantitative and
qualitative approaches in the framework of one study. The proponents of this view
are sometimes called situationalists.
The advocates of the third position regard quantitative and qualitative approaches
both as useful and proper ways of going to study the social world. Although they
see some major differences between quantitative and qualitative research they also
see some important similarities between them and advocate the integrated use of
different methodologies if this can advance our understanding about the
phenomenon under the investigation. The proponents of this position are
sometimes called pragmatists.
All of these three positions bring up some skeptical questions and problems one
needs to address and solve. In this paper I will take the pragmatist position, which
means that I will not question the feasibility of combining quantitative and
qualitative ways of doing research in general. I rather
try to look more closely on problems, which we have to be aware of in the process
of doing so.
Thus, I will not discuss problems, which paradigmatic view brings with it as this
subject has been the focus of many previous papers (including mine from the last
conference in Lahti).
Although the calls for the use of multiple methods in the framework of one study
are maybe even older than the quantitative-qualitative debate, the area of 'how,
when and why different methods might be combined' has got much less attention
than the philosophical aspects of paradigmatic view (Bryman 1988, 155). One can
not say that there is a complete lack of literature concerning different aspects of
combining divergent methodologies. Still most of the literature, which classifies
under the broad area of combining qualitative and quantitative approaches, are
arguments why this integration is possible and needed. On the other hand there is a
considerable number of papers either describing authors' own experiences on
integrating some aspects of quantitative and qualitative methodologies or following
so called 'case law' approach where a number of different experiences are
assembled together and called upon as exemplars one could follow (see for
example Brannen 1992, Brown et al. 1996, Bryman 1988, Caracelli & Greene
1993, Carey 1993, Maxwell et al. 1986).
'get on with the business of attacking our problems with the widest array of conceptual and
methodological tools that we possess and they demand' (Trow 1957, 35; also quoted in
Brewer & Hunter 1989).
Approximately as early Donald Campbell and his co-authors published several
papers where they advocated the use of multitrait-multimethod matrixes and
triangulation of measurement for validation, proposed 'transition experiments' and
quasi-experimental designs (see Campbell 1957, Campbell & Fiske 1959,
Campbell & Stanley 1963, Webb et al. 1966). Although remaining in the
framework of quantitative tradition we can see in these early works the attempt to
advocate the use of multiple methods as well as the possibility to mix some aspects
of quantitative and qualitative methodologies.
Triangulation
Drawing on these ideas Denzin (1978) developed the concept of triangulation - the
term that is probably most widely used to denote any attempt to combine or mix
different methods in a research study. As it often happens, the most widely used
terms tend to be the most overused and abused terms as well, and 'triangulation' is
not an exception here I think. One could draw obvious parallels in how the term's
'paradigm' and 'triangulation' have lost their initial quite narrow and well-defined
meaning and became to denote something general and indefinite.
data triangulation - the use of variety of data sources and data sets in a study. Data
may be both qualitative and quantitative, gathered by different methods or by the
same method from different sources or at different times.
We can see that the concept of triangulation is based on the assumption that by
using several data sources, methods and investigators one can neutralize bias
inherent in one particular data source, investigator or method (Jick 1979). It is
often stressed out that different methods have different weaknesses and strengths
and therefore the main effect triangulation can offer is to overcome the weaknesses
of any single method. Thus, if we use several different methods for investigation of
the phenomenon of our interest and the results provide mutual confirmation we can
be more sure that our results are valid. Within this context, quantitative and
qualitative approaches are usually seen as different ways of studying the same
phenomenon and able to answer the same research questions (Bryman 1988).
On the other hand in the wider framework of integrated use of qualitative and
quantitative approaches the triangulation is offering quite limited possibilities. As
in the case of triangulation the results of different methods are supposed to validate
each other it means that different methods have to be highly independent
throughout the study. This approach excludes the possibility to mix quantitative
and qualitative aspects on different levels of investigation. For example one of the
few books devoted entirely to the problems of combining quantitative and
qualitative approaches Multimethod Research. A Synthesis of Styles by Brewer and
Hunter (1989) is largely constrained by the framework of triangulation although
the authors mention the other possibilities for integration as well. They classify
studies into three categories: a) monomethod studies, b) composite method studies,
which combine some elements of the basic monomethod styles and c) multimethod
studies, which combine the basic styles of research. The authors give their clear
preference to the multimethod designs because according to them composite
methods 'comprise some of the basic methods' sources of sterngth' and they 'fail to
provide the opportunity for triangulated measurement and hypothesis testing, and
the protection against monomethod bias, that the multimethod strategy provides'
(Brewer and Hunter 1989, 81).
There are some methodologists who propose that the combination of various
elements of quantitative and qualitative approaches can offer much wider
possibilities than Brewer and Hunter describe (See for example Brannen 1992,
Bryman 1992, Datta 1994, Patton 1990, Cresswell 1995, Tashakkori and Teddlie
1998). Although these authors see the potential of studies with various mixed
designs, their overall concern seems to be that we still lack a comprehensive
theoretical background for mixed-model designs and therefore mixed-model
studies can seem to many practitioners, consumers and evaluators of the research
as 'mixed-up models' one can not rely on (Datta 1994, 59).
There have been several attempts to clarify the issue and develop taxonomies for
classification of studies combining quantitative and qualitative approaches. For
example Mark and Shotland (1987) introduce in addition to triangulation two other
ways of combining methods to enhance an investigation. They describe
the bracketing model, which says that we should consider the results of different
methods as alternative estimates. Thus, by using methods that are biased in
opposite directions the true value can be bracketed. The third model they describe
is complementary multiplism. Here diverse methods play complementary roles and
offer different viewpoints; together, they provide evidence that is markedly
strengthened.
In one of the most important books on the field Bryman (1988) described in
addition to triangulation ten other ways in which quantitative and qualitative
research have been combined in the research practice. Although it is a quite long
list Bryman conceded that very likely it is not exhaustive (See Table 1). Drawing
on the results of extensive literature review Greene, Caracelli and Graham (1989)
developed a system consisting of five different mixed-method purposes. The
results of their empirical study proved this taxonomy to be exhaustive as it was
possible to match all mixed-method studies, they looked at, to one or more of these
five purposes. I have tried to compare these two lists of purposes for mixed-
method studies in Table 1. We can see that there are some remarkable similarities:
in addition to triangulation the common purposes seem to be development and
expansion. In the former case the results from qualitative research help to inform
the use of quantitative research or vice versa and in the latter the quantitative and
qualitative approaches are used sequentially on different stages of the study. It is
not so easy to draw parallels between other categories as many of Bryman's ways
of integration can serve diverse purposes from complementarity through initiation
to expansion.
Looking for further references to mixed designs we can see that as early as 1980
Patton added a chapter on 'methodological mixes' to his widely used textbook
about qualitative evaluation methods. While also accepting the importance of
triangulation he wrote that the second way of 'achieving methodological
heterogeneity ... is to borrow and combine parts from pure methodological
strategies, thus creating mixed methodological strategies' (Patton 1980, 109, italics
added). Similarly Cresswell describes in his textbook three different designs for
combined use of quantitative and qualitative research: two-phase design,
dominant-less dominant design and mixed-methodology design, where the
researcher 'would mix aspects of qualitative and quantitative paradigm at all or
many methodological steps in the design' (Cresswell 1995, 178).
What we can learn from these different sources is that, as on any developing area,
there is a lack of terminological and even conceptual clarity and coherence: we can
find many different labels for the same ideas; at the same time authors use same
terms in different meanings (See Table 2)., Recently Abbas Tashakkori and
Charles Teddlie (1998) have made an attempt to develop exhaustive taxonomies
for studies combining quantitative and qualitative approaches in different ways.
They have not chosen to proceed from different purposes that combination of
different methods can serve but from the way in which different methods are
used/combined in a particular study. Similarly to Brewer and Hunter (1989) they
divide all studies into three main types: monomethod studies, mixed method
studies, and mixed model studies. Each of these basic types of studies is further
divided into subcategories. The basic types are defined as follows:
Monomethod studies follow in all stages of the inquiry 'one of the predominant
paradigms' (Tashakkori & Teddlie 1998: 17). Thus, they are either purely
quantitative or purely qualitative studies.
Mixed model studies 'combine the qualitative and quantitative approaches within
different phases of the research process' (Op sit: 19). Here qualitative and
quantitative methodologies can be interwoven in different ways. They can be
present as the single application on different stages of the study or they may be
used simultaneously in integrated manner on the same pahse(s) of the inquiry.
Therefore it seems to me that the use of the word 'multi' or 'multiple' would be
more appropriate for this category.
* complementary multiplism
Bryman 1988 quantitative study; ten different ways of methodological hybrids
integration
qualitative study
Patton 1980 quantitative study; triangulation mixed-methodology design
qualitative study
*
* triangulation and bracketing model can be used within the purely quantitative or qualitative studies as well.
In addition to these six mixed types where on every stage of the study there is used
only single approach Tashakkori and Teddlie gave examples of studies with more
complex mixed designs where both quantitative and qualitative approach are used
one the same stage of the inquiry.
The authors claim that proposed three expanded dimensions with relevant
categories are representative of basic steps one has to take in the course of the
research study. However, by studying their definitions I see an important gap in
these dimensions. The first dimension type of investigation is based on the
'distinction between studies with a priori hypotheses (confirmatory investigation)
and those without a priori hypotheses (exploratory investigations)' (Tashakkori &
Teddlie 1998: 53). The second dimension data collection and operations concerns
in the first place 'the form of data', but it also includes the measurement techniques,
methods for establishing the reliability and validity of result as well as sampling
procedures (op sit: 54). My main concern here is that they exclude the overall
design or strategy from their dimensions or put it implicitly into the same
dimension with data collection techniques. I even tend to object the idea that
sampling should be in the same group with data collection techniques, as the
results of my small-scale study showed that there were studies with qualitative
design and sampling, which used mainly quantitative data collection techniques
(Niglas 1999). Thus, I propose that overall design and sampling techniques should
form the separate dimension between the type of the study and the data collection.
If we look at Patton's original classification we see that his first dimension
comprises the overall design of the study, but the weakness of his classification is
that the given categories: naturalistic inquiry versus experimental design are too
restrictive, excluding many design options there are available.
Discussion
In this paper I have given a short overview of the evolution of ideas how
quantitative and qualitative approaches can be combined in the social research to
advance our knowledge about important aspects of our life. We have seen that the
idea to use multiple methods in the framework of one study was proposed already
in the middle of the past century by influential methodologists like Campbell,
Stanley and others. Soon the idea was taken further suggesting that the
combination of quantitative and qualitative research, which were by many
methodologists seen as incommensurable opposites, is not only feasible and
beneficial in solving our puzzles, but can solve some problems the 'pure designs'
can not overcome. Relying mainly on the examples from the research practice
different authors have listed various reasons for combination of quantitative and
qualitative aspects in a single study. More recently the attempts to chart the area by
developing taxonomies for studies combining quantitative and qualitative research
in different ways have been made. In this paper I have described several
possibilities of this classification proposed by different authors and pointed to
some problems related to the definition of dimensions which could be the basis for
classification of mixed designs.
In this short review article it has not been possible to focus on full list of important
issues related with possibilities and ways of combining quantitative and qualitative
research. In spite of their rejection of the incompatibility thesis several authors
have pointed to important problems one has to be aware of and deal with while
combining different elements of quantitative and qualitative approaches (See for
example Bryman 1992). Further, there are strong arguments that it is
oversimplified and even wrong to talk about two clearly distinguishable
approaches as the differences within quantitative and qualitative approaches are not
smaller or less important than between these two categories (Hammersley 1995 a).
On the one hand this argument supports the idea that indeed we can and we do
combine or mix design elements usually connected with divergent approaches, but
on the other hand it questions the basis for taxonomies of mixed designs as
whatever dimensions we take it is not possible to define two clearly separate
categories: quantitative and qualitative. As an example I would mention the
confusion about the methods of manifest content analysis, which have been seen
by some authors as quantitative and by others as qualitative data analysis
techniques.
In summary it has to be said that although some aspects of combined use of
quantitative and qualitative approaches have been studied in more detail (see for
example the work on data analysis strategies for mixed-method studies by
Caracelli and Greene (1993)), there is a need for further research to clarify the
implications of various mixed designs to the research results.
References
Brown, S., Riddell, S. & Duffield, J. (1996). Possibilities and problems of small-
scale studies to unpack the findings of large-scale studies of school effectiveness.
In Gray, J., Reynolds, D., Fitz-Gibon, C. & Jesson, D. (eds) Merging traditions:
The future of research on school effectiveness and school improvement. London:
Cassell.
Campbell, D.T. & Fiske, D.W. (1959) Convergent and discriminant validation by
the multitrait-multimethod matrix. Psychological Bulletin, 65, pp. 81-105.
Caracelli, V.W. & Greene, J.C. (1993) Data analysis strategies for mixed-method
evaluation designs. Educational Evaluation and Policy Analysis, 15(2), pp. 195-
207.
Greene, J.C., Caracelli, V.J. & Graham, W.F. (1989) Toward a conceptual
framework for mixed-method evaluation designs. Educational Evaluation and
Policy Analysis, 11, pp. 255-274.
Hammersley, M. (1992a) The Paradigm Wars: Reports From the Front. British
Journal of Sociology of Education, 13(1), pp.131-143.
Mark, M.M. & Shotland, R.L. (1987) Alternative models for the use of multiple
methods. In Mark, M.M. & Shotland, R.L. (eds) Multiple methods in program
evaluation: New Directions for Program Evaluation 35, pp. 95-100. San
Francisco: Jossey-Bass.
Maxwell, J.A., Bashook, P.G. & Sandlow, C.J. (1986) Combining ethnographic
and experimental methods in educational evaluation. In Fetterman, D.M. & Pitman
M.A. (eds), Educational evaluation: Ethnography in theory, practice, and politics.
Bewerly Hills, CA: Sage.
Patton, M.Q. (1980) Qualitative evaluation and research methods. Newbury Park,
CA:Sage.
Reichardt, C.S. & Cook, T.D. (1979) Beyond Qualitative versus Quantitative
Methods. In Cook, T.D. & Reichardt, C.S. (eds) Qualitative and Quantitative
Methods in Evaluation Research. Beverly Hills, Calif.: Sage, pp. 7-32.
Reichardt, C.S. & Rallis, S.F. (1994) Qualitative and quantitative inquiries are not
incompatible: A call for a new partnership. In Reichardt, C.S. & Rallis, S.F.
(eds) The qualitative-quantitative debate: New perspectives. San Francisco:
Jossey-Bass.
Thompson, D. (ed) (1996) The Oxford Dictionary of Current English (revised 2nd
ed). Oxford University Press.
Webb, E.J., Campbell, D.T. Schwartz, R.D. & Sechrest, L. (1966) Unrobustive
measures.: Nonreactive research in the social sciences. Chicago: Rand McNally.
Types
The problem with relying on just one option is to do with bias. There are
several types of bias encountered in research, and triangulation can help
with most of them.
Sources
Carvalho, S. and White, H. (1997). Combining the quantitative and
qualitative approaches to poverty measurement and analysis: The practice
and the potential. World Bank Technical Paper 366. Washington, D.C.:
World Bank
https://www.betterevaluation.org/en/evaluation-options/triangulation