“Triangulation” in social science is the use of 2 or more sources of data, or 2 or more methods of data collection (mixed methods, multiple methods), in order to check/verify results; in order to increase the credibility of results (Triangulation (social science)).
Although there’s general agreement about the advantage of triangulation over mono-method research, we need to recognize that its advantage covers fundamental flaws in social science methodology. The first step in uncovering these flaws is to recognize that triangulation is not a topic in natural science. Physicists, chemists, and other natural scientists increase confidence in the accuracy of their data, not by triangulation, but by replications of agreed-upon procedures, such as observation and experimentation, that they, again, agree generate accurate or correct data. Chemists, for example, increase their confidence in, and the accuracy or reliability of, their data by performing multiple experiments. Astronomers become more confident in the accuracy of their data, not by one look at the sky, but via multiple observations.
Social scientists cannot optimize their confidence in data obtained from a single source (say, from documents), or from one method (e.g., observation), or even by replications thereof, and must employ triangulation because social scientists don’t agree on which methods generate accurate data and, thus, don’t agree on what constitutes accurate data. Moreover, social scientists don’t agree on the characteristics of investigated phenomena; that is, they differ in their views or understandings of the (a) characteristics of human beings and the (b) characteristics of the context/environment of humans. Some have a behavioral view of humans; others (the vast majority) are non-behaviorists. Concerning the context/environment of humans, some social scientists interpret it from a system analysis perspective, emphasizing the interdependence, dependence, etc., of the parts of the system, whereas others have a non-system analysis view that affirms the existence or possibility of the independence or autonomy of investigated phenomena.
Given this methodological muddle and these disagreements, triangulation, rather than mono-method research, is the best procedure. However, what’s more important for strengthening social science methodology is (1) to recognize that social scientists need triangulation only because they don’t agree about which procedures produce reliable data and, then, (2) to eliminate the need for triangulation by establishing agreement among social scientists (as there is among natural scientists) about which methods produce reliable data. Bringing about this agreement requires “the ingenuity of the shared enterprise“; a concerted effort throughout social science that moves us, not just beyond mono-method research but, also, beyond triangulation to methodological agreement. Individual and institutional contributions to methodological agreement in social science include Eugene Webb, et al., Unobtrusive Measures, Gary King and Harvard’s Institute for Quantitative Social Science, Rein Taagepera’s Making Social Sciences More Scientific: The Need for Predictive Models, my methods course: How to Find Out What’s Really Going On: Observation, Experiments, Multiple Sources (of non-askng data), Models, Content Analysis, and Comparison (of non-asking data), and I’m sure there are others. (Let me know in the Comment section below if you have additions to this list.)
Along the way to methodological agreement, we’ll improve our procedures for obtaining reliable data (e.g., by greater use of IT) and, moreover, our agreement and advances, as they make us more scientific, will drive us to also agree about the characteristics of human beings (the behavioral approach will prevail), as well as to agree about the context or environment of humans (system analysis will prevail).