470 QUALITATIVE INQUIRY / December 1999
extremes), but want to use it to illustrate the limits of any approach that expects research practice to conform precisely to a philosophical position.
Subtle realism involves maintaining a view of language as both constructing new worlds and as referring to a reality outside the text, a means of communicating past experience as well as imagining new experiences. Hammersley (1992, 1995) presents this as an adequately grounded place for social researchers seeking a middle way between the various paradigm positions that are nowadays available. Like analytic realism (Altheide & Johnson, 1994), Kantian soft or “transcendental” idealism, and critical realism (Bhaskar, 1989), it is a marker of an approach to social research that takes the view that, although we always perceive the world from a particular viewpoint, the world acts back on us to constrain the points of view that are possible. The researcher treading this middle way is continually aware of the somewhat constructed nature of research but avoids the wholesale application of constructivism to his or her own practice, which would result in a descent into nihilism. Research, then, constructs “transitive objects” such as the concepts of social science, to represent the real (Williams & May, 1996, p. 85). Knowledge is always mediated by preexisting ideas and values, whether this is acknowledged by researchers or not. Yet, some accounts are more plausible than others, and human communities in practice have created reasonably firm grounds on which plausibility can be judged, whether or not these grounds can be supported in some ultimate sense by means of philosophical reasoning. Judgments about the plausibility of research accounts inevitably involve a temporary subscription to the view that language is referential to a reality outside the text. This is a long way from a simple correspondence theory of truth, but it contains elements of this. Neither does it claim that truth solely lies in the consistency of claims with some other set of claims, though this can legitimately be an element in judging truth claims. It involves opposition to the pure constructivist view that states there is no possibility of knowing a real world that exists separately from language.
At the heart of the advocacy of subtle realism lies the idea of a research community with agreed standards of judgment for the plausibility, credibility, and relevance of research reports. Distinguishing claims from evidence, providing the strongest evidence for more important claims, and exposing the judgments of the researcher for readers to scrutinize are all methods for addressing the standards applied by a community of critical peers. In arguing for this, Hammersley (1992) pursues an argument similar to that of Popper (1972), who claimed the authority of an imagined “third world” of objective knowledge, humanly constructed but, by virtue of being a joint endeavor of a community of scientists, having an existence independent of the biographies of individual scientists. Hammersley (1992) is also similar to Popper (1963) in advocating a fallibilistic approach, regarding “truths” as provisional until
Seale / QUALITATIVE RESEARCH 471
there is good reason for contradictory versions to gain support. Hammersley is therefore firmly in a postpositivist camp.
As is well known from the criticisms of Popper that have emanated from radical epistemological and political positions, reliance on norms of communal assessment has the potential to support a rather conservative approach. We can observe that this community of researchers is not in fact an imagined thing of the mind, but a reality. Particular people do concrete things in the world and call them research. These people come from particular cultural backgrounds and bring specific, exclusive prejudices to bear in the standards that they maintain. In practice, the social research community is no different from the rest of society in its divisions of status and power, acting at times to oppress and silence particular groups who are unable to influence the discourses of social research (Harding, 1986). Hammersley’s stress on whether findings are consistent with knowledge that is currently accepted in the relevant research community (“plausibility”) initially looks rather dubious in this light. Against this, though, the advocate of subtle realism might point to the role of evidence in testing theories for both credibility and plausibility, which exerts a persuasive force on the research community and can result in revision of accepted wisdom and the eventual overthrow of dominant paradigms.
Yet, this is itself assailable, as it rests on assumptions about the evidentiary basis of what constitutes evidence. How should we make contact with an external reality that affirms or disaffirms claims? Is not all observation fundamentally driven by preexisting theoretical assumptions? To present subtle realism as a foundational basis for social research practice seems inadequate in light of these questions. Conventionally, there can be a turn to further philosophical work at this point, perhaps to constructivism or postmodernism, often claimed to be epistemologically nonfoundational, but nevertheless presented by some (for example, Denzin, 1997; Dickens & Fontana, 1994) as a new rationale for research practice, suggesting a foundationalist habit of thought with which I believe researchers should break.
The widespread appeal of postmodern, political, and constructivist conceptions of research is based on some fundamental dissatisfactions with the scientific world view. Quality does matter in qualitative research, but the modernist headings of validity and reliability no longer seem adequate to encapsulate the range of issues that a concern for quality must raise. The constructivist critique of criteriology has led us to see that “quality” is a somewhat elusive phenomenon that cannot be prespecified by methodological rules, though their reconstitution as “guidelines,” to be followed with intelligence and knowledge of the particular research context, may assist us in moving toward good quality work. A major threat to quality is in fact the idea that research must be carried out under the burden of fulfilling some
472 QUALITATIVE INQUIRY / December 1999
philosophical or methodological scheme. Practicing social researchers can learn to do good work from a variety of examples, done within different “moments,” without needing to resolve methodological disputes before beginning their work. At the same time, the quality of qualitative research is enhanced if researchers engage with philosophical and methodological debate, so that the pursuit of quality becomes a “fertile obsession” as methodological awareness develops and feeds into practice (Lather, 1993).
The idea of a self-critical research community acting together to produce positive knowledge for the benefit of others retains its appeal for many researchers. The continuing desire to participate in a shared language, constructing and negotiating standards for judging quality, incorporating political and cultural differences, always involves an act of trust in the judgments of others, though this can be made easier by the application of certain methodological procedures. These procedures (discussed in more depth in Seale, 1999), are based on this view of a research community existing as a key audience for social researchers concerned about the quality of their efforts. They include techniques such as the peer auditing described by Lincoln and Guba (1985). Acceptance of the researcher’s case can then partly depend on the capacity of the researcher to expose to a critical readership the judgments and methodological decisions made in the course of a research study (Swanborn, 1996).
TRIANGULATION: A CASE
STUDY OF A CRAFT SKILL
To illustrate an approach to research practice that cuts through inconclusive methodological disputes, retaining a conception of research as primarily a craft skill, I shall discuss a particular one of these skills. Triangulation describes a set of techniques that arose initially within a crudely realist paradigm. Attempts have been made to restrict its use to this paradigm by people who mistakenly believe in inevitable logical connections between paradigm positions and techniques. More enlightened methodologists have perceived that it has a place within a variety of paradigms. The next step in this logic, which I believe researchers should take, is to claim it as a valuable craft skill, relatively autonomous from any paradigm position.
The idea of triangulation derives from discussions of measurement validity by quantitative methodologists working with crudely realist and empiricist assumptions. Campbell and Fiske (1959) argued that “In contrast with the single operationalism now dominant in psychology, we are advocating . . . a methodological triangulation” (p. 101, emphasis in original), and proceeded to outline their ideas for the convergent and discriminant validation of measurement instruments. Subsequently, Webb, Campbell, Schwartz, and Sechrest (1966) used the idea to advocate multiple operationalism, the use of
Seale / QUALITATIVE RESEARCH 473
several methods at once so that the biases of any one method might be canceled out by those of others. Its use in qualitative research, though, was first advocated and then popularized by Denzin (1970) whose textbook has been through several editions (1978, 1989) in which the original concept was modified. Other textbook definitions at times contain distant echoes of this background in the quantitative research tradition, as where Hammersley and Atkinson (1983) describe it as a method whereby “links between concepts and indicators are checked by recourse to other indicators” (p. 199). The term itself is designed to evoke an analogy with surveying or navigation, in which people discover their position on a map by taking bearings on two landmarks, lines from which will intersect at the observer’s position. If only one landmark were taken, the observer would only know that they were situated somewhere along a line. Triangulation used in this way assumes a single fixed reality that can be known objectively through the use of multiple methods of social research (Blaikie, 1991). Many might feel that it is therefore a technique impossible to employ without also taking on modernist philosophical commitments within a positivist, or at least postpositivist, paradigm together with a commitment to constructing a single true version.
Cicourel (1964, 1974) offers the most extreme vantage point from which to view triangulation from a different paradigm position. His own critique of the technique is typically paradoxical, in that he proceeds by enthusiastically advocating the advantages of “indefinite triangulation” (Cicourel, 1974, p. 124). This rhetorical ploy (that I think could usefully be read now as a joke, though this was probably not Cicourel’s intention) proceeds by showing that what he means by this is in fact the precise antithesis to the consensus on truth sought in conventional triangulation. His illustration comes from his own practice.
The triangulation procedure varies with the research problem. When gathering information on language acquisition in the home setting we left a tape recorder for about one hour during lunch. A transcription of the tape was done by a typist who had been instructed to render a verbatim record. Then the transcript, the first version of this scene, was read by the mother while she listened to the tape; her comments produced another version of the interaction. The typist was next asked to listen again to the tape and to describe what she thought was “going on,” correcting her original transcript as she deemed necessary. In this elaboration and correction a different version of the scene was always produced. My phonetic transcription of the tapes created still another version. . . . The reader could now say that we should have simply combined the different versions to produce the “best” one possible, but the point is that different versions could have been produced indefinitely by simply hiring different typists and providing the mother with different transcripts. (Cicourel, 1974, p. 124)
This is an amusing little demonstration of the constructivist objection to realist tendencies in discussions of triangulation. As with postmodern views, it suggests that every reading of a text is likely to produce a new interpretation, with no version assuming privileged status. Cicourel’s example leads Blaikie (1991) to claim that triangulation therefore only makes sense from within a
474 QUALITATIVE INQUIRY / December 1999
positivist framework. Blaikie fails to see any irony in Cicourel’s example (“But the question arises as to why it should be called triangulation” [Blaikie, 1991, p. 130]) and argues that the technique “has no relevance for genuine interpretivists and ethnomethodologists” (Blaikie, 1991, p. 131) because it necessarily involves subscription to inappropriate ontological and epistemological positions. I argue, by contrast, that researchers should question this claim that there is a necessary connection.
Bloor’s (1997) objection to triangulation leads to a further questioning of its supposedly inevitable philosophical connotations, though it also offers an opportunity to locate this skill within a subtle realist paradigm. Bloor (1997) says that even if all the different methods in a methodological triangulation exercise converge on the same thing, apparently agreeing with each other, how can we know that they are correct? Perhaps some hitherto unthought-of method would reveal something different. In fact, this problem is analogous to that of induction: How can we reliably reason on the evidence of past experience that the sun will rise tomorrow? Logically of course, we cannot. Taken at this level, the objection to triangulation as a validation exercise is also unanswerable. Yet, we operate in the world all the time on the basis of what it is plausible to believe, and it will do us little good to assume that the sun will not rise tomorrow. A pragmatic, subtle realist might answer Bloor by saying that triangulation exercises can add to the credibility of a particular account as a part of a fallibilistic research strategy in which evidence is sought for central claims. Thus, we can move from a crude to a subtle realist paradigm as a potential set of background assumptions for triangulation. Can we go further toward Cicourel’s view?
Silverman (1993) presents an argument that moves triangulation toward a constructivist paradigm. At first, he makes the subtle realist point that triangulation exercises can deepen understanding as a part of a fallibilistic approach to fieldwork, although being themselves no guarantee of validity. In a similar vein, Cain and Finch (1981) argue that multiplication of methods can help deepen understanding of different aspects of an issue. Dingwall (1997) begins the move toward constructivism by saying that triangulation offers a way of explaining how accounts and actions in one setting are influenced or constrained by those in another. Silverman (1993) eventually has no problem with this use of triangulation, finally saying that it can help “to address the situated work of accounts” rather than “using one account to undercut the other” (p. 158). This version of triangulation, then, which is now very close to Cicourel’s view, gets away from the idea of convergence on a fixed point and accepts a view of research as revealing multiple constructed realities, something that triangulation, now conceived as the revelation of difference, is well suited to expose.
Flick (1992, 1998) completes this move toward what might be called a soft constructivist version of triangulation, deriving this from a study in which a