Mixed Methods,Triangulation,and Causal Explanation

Disjunctive Mixed Methods/Triangulation ¹

Causal claims take the general form “something makes something else happen” (Searle, 1984, p. 65). In social research, there are two ways to conceive of the relationship between the something and the something else: mechanical and agential. ² Mechanical causation (M-causation) construes causal explanation as identifying and accounting for ordered processes of human behavior on the model of the natural sciences. M-causation is typically associated with quantitative-experimental methods and is adopted by mixed methods experimentalism as the only legitimate conception of causal explanation in social research. ³ In this view, there is a “unity of science.” Explaining human behavior is like explaining the behavior of machines. Or at least the difference between humans and machines is one of degree, not kind. The complexities introduced by human agency are interpreted as “noise” that requires increased “error limits” (NRC, 2002) and, in turn, render explaining human behavior more like explaining the behavior of somewhat erratic machines than smooth-running ones. This fits quite logically with adopting randomized experiments as the “ideal” method of investigating causal relationships (NRC, 2002), for they provide the most effective way to evenly disperse the noise that you can then divide through by and cancel out.

Agential causation (A-causation) construes causal explanation as identifying and accounting for ordered processes of human behavior in terms of norm-governed institutions and practices. A-causation is typically associated with qualitative-interpretive research methods and has roots that can be traced as far back as Aristotle’s conception of phronesis (e.g., Flyvbjerg, 2001). But despite this long history, it is much less familiar than M-causation and, for that reason, requires considerable elaboration.

In contemporary philosophical work, A-causation is grounded in the concept of “intentionality” (e.g., Dennett, 1987; Searle, 1984, 1995). Intentionality includes conscious intentions to act but also a much broader domain of mental concepts that have the characteristic of “aboutness,” for example, beliefs, doubts, and knowledge. Fears and worries are also intentional. Intentional concepts are of particular interest in social research because they go into descriptions of the actions people perform, as distinct from the mere movements of physical bodies. Consider the assertion “Jodi’s arm went up.” If this happened while Jodi sat watching TV, the result of a peculiar kind of spasm to which she was prone, it would be a movement (an instance of M-causation) not explicable in terms of intentionality. On the other hand, if Jodi’s arm went up (the same movement) in the context of a discussion in an American school classroom, we would take this to be an action in which Jodi was seeking to be recognized by the teacher. And if we were to explain things in this way to a visitor unfamiliar with the conventional practice of hand raising in American classrooms, it would provide the visitor with an A-explanation of Jodi’s behavior. In general, A-explanations appeal to the role intentionality plays in human behavior. This form of explanation becomes quite complicated when fully fleshed out as a framework for causal explanation.

John Searle (1995) develops such a framework in which three concepts are of particular importance: collective intentionality, social facts, and the Background. When human beings cooperate in the pursuit of goals, they exhibit collective intentionality, through which they construct a special class of social facts that, unlike the brute facts of the physical world, wouldn’t exist but for the activities of human beings. Take money, Searle’s most perspicuous example. That money is a human construction with no independent existence is apparent. Indeed, nowadays it is only loosely linked to physical tokens such as coins, paper currency, and checks; it exists and is exchanged primarily in cyberspace. Money’s value has nothing to do with the physical form it takes but is to be found in our collective acceptance of and actions consistent with the rules of monetary exchange.

Money is a special type of social fact, an institutional fact, in which status functions are represented by the formula “X counts as Y in C”—for example, “A five-dollar bill (X) counts as (has the status of) money (Y) in the United States (C).” Status functions also apply to institutional roles, for example, teacher, physician, and barber. Along with these roles go certain kinds of powers and responsibilities, which are created by the assignment of the status functions. Less formalized social facts, like the roles associated with being a member of Greenpeace, for example, function in similar ways to institutional facts in assigning status functions, powers, and responsibilities.

Social and institutional facts are underlain by collective intentionality and are thus human constructions in the sense that they wouldn’t exist but for the activities of human beings. But it is important to observe that the concept of collective intentionality does not assume that agents always or even usually consciously follow the rules governing status functions or even know that such things exist. ⁴ Consider 12-year-old Maria buying lunch in her school cafeteria. She need have no conception of money as a status function to successfully act in accord with what it requires.

We might say that in competently using money, Maria is tracking the rules rather than following them. Searle (1995) introduces the “Background” to help account for this phenomenon. He describes the Background as “the set of nonintentional or preintentional capacities that enable intentional states to function” (p. 129). These capacities include abilities, dispositions, and know-hows. They enable things such as linguistic and perceptual interpretation, they structure motivation, and they dispose persons to certain kinds of behavior. Searle explains,

One develops skills and abilities that are, so to speak, functionally equivalent to the system of rules, without actually containing any representations or internalizations of those rules.

There is a parallelism between the functional structure of the Background and the intentional structure of the social phenomena to which the Background capacities relate. That strict parallelism gives the illusion that the person who is able to deal with money, cope with society, and speak a language must be [consciously or] unconsciously following rules. (p. 142)

In many situations, Searle says, “we just know what to do,” and the idea of following rules does not apply. We do follow rules sometimes—as in making a legal argument, registering to vote, and so forth—but the applicable rules are never self-interpreting and are never exhaustive. Thus, even consciously following rules calls for the exercise of interpretive and creative capacities associated with the Background.

In terms of the description of A-explanation provided at the outset—identifying and accounting for ordered processes of human behavior in terms of rule-governed institutions and practices—Searle’s framework (1) identifies the social facts (including institutional facts) and associated status functions constructed by given social groups and (2) explicates how members of that group (a) follow or (b) track the associated rules.

Under Searle’s account, humans typically have a quite limited say in what the social facts and rules are, for they are born into a social life circumscribed by these facts and rules and by and large just catch on to how to behave. Because they are shaped by external influences in this way, A-explanation seems subsumable under M-explanation.

But this conclusion is mistaken. That humans are shaped by external influences does not preclude that there is an arena in which they play an active role in shaping their own lives. This varies significantly among individuals, of course, for the degree to which they have the opportunity to shape their own lives depends on the kind of political regime they inhabit, their social position, the dispositions and skills they have, and the opportunities they are given to develop. But more generally, the social shaping of individuals via systems of collective intentionality is quite different from the kind of mechanical shaping that characterizes how the weather interacts with the type of rock in a given region to shape a mountainside. A mountainside does not behave wrongly if it fails to erode as predicted. And, assuming the data were accurate, the explanatory framework of laws that led to the inaccurate prediction would have to be revised. In contrast, social shaping is normative, and its laws do not have to be revised when violated. An embezzler does behave wrongly and, moreover, the occurrence of his transgression does not require that laws applying to embezzlement be revised. The laws remain in place, to be obeyed in response to the threat of moral criticism, prison time, and the like. Natural objects obey laws only in a metaphorical sense.

Thinkers such as Searle deny the unity of science and turn mixed methods experimentalism on its head. Observed regularities in human behavior and the relationships among them, frequent ingredients of M-explanation, do not provide causal explanations in the social realm. They are merely flags. For example, given that we know something about humans, that cars occupied by them regularly come to rest when traffic lights are red indicates something causal is likely going on. (And if it were in a future in which cars are designed to automatically respond to traffic lights, the regularity would flag M-causation.) But the observed regularity by itself does not provide an explanation of why humans stop when traffic lights are red. Stopping is an action that can only be understood in terms of traffic laws and humans’ reasons for obeying them, and it is this that underlies and explains the observed regularity. This position—mixed methods interpretivism (Howe, 2004)—also embraces disjunctive conceptions of mixed methods and triangulation, but it is quantitative-experimental methods that are relegated to description and exploration and qualitative-interpretive methods that do the work of providing causal explanations.

Conjunctive Mixed Methods/Triangulation

In exploring conjunctive mixed methods/triangulation, it is important not to conflate two levels of “mixed methods.” Critics of a conjunctive form of mixed methods research have long conceded that their concerns do not apply to “techniques and procedures” (e.g., Smith & Heshusius, 1986), say triangulating with quantitative and qualitative data. They object, instead, to the idea that triangulation is coherent at the level of broader social research “paradigms,” particularly positivism versus interpretivism. I have spilled plenty of ink trying to dispel the “incompatibility thesis”: Remove positivism from the scene—long since abandoned in philosophy—and you remove the grounds for paradigm incompatibility (Howe, 2003). Nonetheless, there remains a question of how to think about combining the alternative conceptions of causation, even if both may be accommodated within a single pragmatic paradigm (Howe, 1988, 2003).

Qualitative and quantitative methods can be used to conjunctively triangulate on A-causation. For example, a quantitative survey instrument may be used to ascertain the reasons people give for voting for or against given policies in order to flag the A-causal relationship between these beliefs and their voting behavior (Moses, et al., 2010). These data may be compared with, and converge with or diverge from, data collected via face-to-face interviews.

On the flipside, qualitative methods and quantitative methods can also be used to conjunctively triangulate on M-causation. For example, physicians interview patients regarding their symptoms to flag the M-causal process underlying the process of disease. These data may then be compared with, and converge with or diverge from, the data obtained from biomedical tests.

These examples illustrate that at the techniques and procedures level, there is no barrier to conjunctively triangulating with qualitative and quantitative methods for questions pertaining to either M-causation or A-causation. But conjunctive triangulation must be viewed differently at the level of causal explanation. As indicated previously, there is quite a difference between an M-causal explanation of the shaping of a mountain by the elements and an A-causal explanation of the shaping of human identities and capacities and their performance of actions in accordance with norm-governed practices. In social research, A-causation is fundamental and can be triangulated with M-causation only in the “holistic” sense suggested by Mathison (1988), in which they are interwoven but addressed to distinct causal processes.

In some studies, A-causation is the primary focus of investigation. For example, in Shirley Brice Heath’s (1983) Ways With Words, the something else of depressed academic performance of African American children is made to happen by the something of norm-governed linguistic practice. In particular, depressed achievement is accounted for in terms of the mismatch between the appropriate and competent use of language between African American homes and White-dominated schools.

In contrast, in other studies, A-causation may be submerged beneath an apparent investigation of M-causation—call it quasi M-causation—exemplifying Searle’s claim that in social research observed regularities in behavior serve only to flag causation. Consider the vaunted student/teacher achievement ratio (STAR) study, a randomized experiment establishing that reduction of K-3 class size to 15 results in higher achievement. It is quite clear that reduced class size per se does not cause increased achievement. Consider how reducing the number of gas molecules in a given container increases the opportunity of two molecules to spend more time in unobstructed proximity with one another. Class size reduction results in the same kind of thing, mechanically speaking. But increasing the opportunity for individual students to spend more time in unobstructed proximity to teachers by itself would do nothing to increase achievement. Rather, teachers must take this opportunity to provide more guidance to individual students on how to do academic work appropriately and competently (“Class Size,” 2011).

Randomized experiments in the social sciences can be very useful in addressing important questions regarding what makes for effective practice. This includes STAR. But this should not obscure the fact that they implicitly incorporate a ceteris paribus clause regarding agency, which renders them much less of a methodological magic bullet than they are so often touted as being (Howe, 2004). The ceteris paribus clause limits their generalizability across both place and time. The STAR finding that smaller class size yields higher achievement didn’t travel well from Tennessee to California (CSR Research Consortium, n.d.), in large part because the kind of interactions between students and teachers present in the STAR study were not replicated. Moreover, even within a given place, the generalizations of social research “decay” (Cronbach, 1975). An important reason for this is what the philosopher Ian Hacking (1999) calls “looping”: humans exercise agency in taking up the finding of social research and sometimes altering their practices in response. Consider the disruptive effects of care theorists such as Carol Gilligan (1982) on the thinking of both women and men. The dissemination of the findings of Gilligan, in which she characterized women as tending toward being motivated by preserving concrete relationships and men as tending toward following the dictates of formal rules, no doubt helped spur change in gender roles.

Finally, A-causation may be interwoven with genuine M-causation. For example, the ingestion of lead paint by children M-causes them to suffer neurological damage, which, in turn, results in effects at the A-causal level of lowered academic achievement and antisocial behavior. In a more complex chain, segregation of African Americans A-causes stigmatization and racial discrimination, which, in turn, A-causes material inequality. Material inequality includes the presence of leaded paint and other environmental toxins that when ingested or inhaled by African American children, M-causes them to suffer neurological damage. This, in turn, results in effects at the A-causal level of lowered academic achievement and antisocial behavior. Lowered academic achievement and antisocial behavior, in turn, reinforce and exacerbate stigmatization, discrimination, segregation via A-causation, and so on. ⁵ Though coherent and indispensable in gaining an understanding of all the dimensions of human behaviors, models that incorporate both A-causation and M-causation possess a discontinuity at the points where these two conceptions of causation make contact.

Conclusion

My primary objective in this article has been to investigate the relationships among mixed methods, triangulation, and causal explanation. I have reached the conclusion that no barriers exist to triangulate both conjunctively and disjunctively at the level of quantitative and qualitative techniques and procedures with respect to either A-causal questions or M-causal questions taken one at a time. However, things are more complicated at the level of more comprehensive causal explanation. First, what is so often equated with the concept of causation to be employed in the explanation of human behavior, as in mixed methods experimentalism, is quasi M-causation that flags underlying A-causation. Second, because A-causal explanations and genuine M-causal explanations preserve their distinctiveness within explanatory frameworks that incorporate them both, such frameworks possess a kind of discontinuity that can only be partly removed by moving to a higher or holistic level of integration.