Mechanisms in the social sciences: the quest for the paradigmatic success
Department of philosophy
Université du Québec à Montréal
The work of Doug McAdam, Sidney Tarrow, and Charles Tilly presents probably the most ambitious application of mechanistic explanations in the social sciences. Over the years, their use and interpretation of mechanisms have come under the criticism of both practitioners and philosophers of the social sciences. Being incomplete, their mechanistic accounts would be hardly distinguishable from storytelling and should be rather understood in heuristic terms. This article reviews the most important criticisms and argues that, although they are for the most justified, they can be circumvented by clarifying the explanandum of mechanistic analysis and by putting realistic demands on what should be included in the description of a social mechanism.
Keywords: Explanation, mechanism; Doug McAdam; Sidney Tarrow; Charles Tilly;
Mechanisms in the social sciences: the quest for the paradigmatic success
The concept of “mechanism” has probably been the single most successful in the philosophy of science, and a fortiori of the special sciences, of the last two decades. Although a variety of mechanistic approaches have currency, they all show dissatisfaction with purely statistical and correlational methods and take “mechanisms” to introduce explicit causal considerations into scientific accounts (Machamer et al. 2000; Bunge 2004; Hedström 2005; Craver 2007; Elster 2007).
As mechanistic approaches gain wider support in the philosophy of the special sciences, new questions arise about how they can help to guide particular research programs and explanations. How can mechanisms be identified or measured? What are the boundaries of mechanisms and what must be included in them? How detailed must the description of mechanisms be and where do mechanistic approaches stand in the debate on reductionism and pluralism?
One way to understand the norms that must guide the search for mechanisms is to look at paradigmatic examples of successful explanations. In the case of neuroscience, for instance, Craver (2007: viii-xi, 65-72, 236-246) has argued that the discovery and description of the mechanism of Long-Term Potentiation—that generates changes in the strength of synapses—was
such a paradigmatic success. It is, Craver contends, because neuroscientists have described all the relevant components, activities, and organizational patterns that exhibit the phenomenon.
Is it possible to identify for the social sciences such successful instances of mechanistic explanations? This question stands in the background of the recent attention given to the
empirical and theoretical work of Doug McAdam, Sidney Tarrow, and Charles Tilly (McAdam et al. 2001, 2008). These authors did not only argue explicitly in favor of mechanisms, but also advanced an impressive mechanistic program for the study of what they call “contentious politics.” This program, by the variety of phenomena that it addresses (social movements,
revolutions, wars, nationalistic movements, etc.), is beyond doubt the most ambitious application of the mechanistic approach in the social sciences. Put forward by some of the most talented and influential social researchers, it is likely to help to understand the potential and limitations of mechanisms, as well as the norms that should guide their use. The extend to which the framework advanced McAdam, Tarrow, and Tilly (henceforth MTT) allows for the successful description of mechanisms has been the object of much discussion since the publication of their book Dynamics
of Contention in 2001.
Several practitioners of social movements and philosophers of the social sciences have discussed and criticized MTT‟s use and interpretation of mechanisms (especially, Koopmans 2003; Oliver 2003; Rucht 2003; Norkus 2005; Earl 2008; Falleti and Lynch 2008; Lichbach 2008; Demetriou 2009). This article reviews the most important criticisms and argues that, although they are for the most justified, they reveal no fatal flaw in the framework.
I begin with a brief overview of MTT‟s interpretation of mechanisms (section 2) and of the main dissatisfactions of commentators with it (section 3). I then present Norkus‟ suggestion
(2005) that MTT‟s approach faces a “dilemma of specification” that blurs the distinction between storytelling and mechanistic accounts of episodes of contention (section 4), as well as Demetriou‟s proposal (2009) to understand MTT‟s mechanisms in terms of a heuristic (section 5).
I argue that these commentators identify real problems in MTT‟s application of mechanism, but that the reality of mechanisms and the distinction with storytelling can be maintained as long as explanandum phenomena are well specified (section 6). To support my point, I take the example
of Tilly‟s later work on Democracy (2007) and show that it withstands relatively well the
criticisms leveled against Dynamics of Contention (2001) (section 7). I finish by discussing one
frustration that might remain with Tilly‟s work, namely, the fact that it does not provide a complete description of the mechanism of democratization (section 8). I argue that this raises a larger problem for mechanistic explanations and that realistic demands must be put on what should be included in a causal mechanism.
2. Mechanisms of contention
In their book Dynamics of Contention (DOC), MTT propose a new framework for the study of
social movements. The classic approach, they argue, suffer from many limitations. The most significant is probably that it is too static. It focuses on structural, cultural, phenomenological, and rational factors, without understanding the properly dynamic character of social movements. It explains mobilization, for instance, by referring to the interests or the identity of agents, and not in terms of alterations in people‟s interest and identity. The second most important complaint of MTT concerns the “compartmentalization” of research on social movements, revolutions, nationalisms, wars, etc. The division of research in subfields is unfortunate because these phenomena are all instances as what they call “episodes of contentious politics.”
DOC offers an answer to these two limitations: different types of episodes of contention must be studied in the same dynamic framework. This framework, they argue, must be centered on two related concepts: mechanisms and processes. They take mechanisms to refer to “a
delimited class of events that alter relations among specified sets of elements in identical of closely similar ways over a variety of situations.” Processes, for their part, are defined as “regular
sequences of such mechanisms that produce similar (generally more complex and contingent) transformations of those elements.” (McAdam et al. 2001: 24)
Three things can be noted about their approach. The first is that the mechanistic approach pretends to be explanatory. Mechanisms are the causes of processes, and processes cannot be explained without reference to mechanisms. The approach is thus intended to go beyond correlations. The second is that mechanisms are not events but sorts of events, which are causally
relevant to the production of a variety of processes and episodes. There is thus an ambition to
generalize and to go beyond particular historical accounts. Finally, the distinction between mechanisms and process is a relative one. The same transformation can be presented as a mechanism or a process depending on the focus of the explanation.
DOC applies the concepts of processes and mechanisms to a large number of events and transformations, of which we can give some examples.
- Examples of processes: annexation, contention, democratization, diffusion,
independence, integration of trust networks into public politics, liberalization,
mobilization, negotiation, radicalization, reform, regime defection, revolution, scale
shift, secularization, state making, transformation of identities, unification.
- Examples of mechanisms: brokerage, category formation, certification, competition,
consensus, escalation, identity shift, mobilization, radicalization, social appropriation,
suddenly imposed grievances, urbanization.
As we can see, some changes are qualified both as mechanisms and processes in the book: mobilization, radicalization, identity shift, etc. MTT (2001) fully acknowledge this fact in discussing one of their favorite changes, “brokerage,” which refers to the establishment of a link between previously unconnected groups or agents: “It is arbitrary, for example, whether we call
brokerage a mechanism, a family of mechanisms, or a process. In this book, we generally call it a mechanism to emphasize its recurrent features.”
The fact of qualifying “brokerage” both as a mechanism and as a process does not pose a problem per se. But the fact that it is called a mechanism because of “its recurrent features,” as MTT argue, is more puzzling. Indeed, MTT define processes as “regular sequences of mechanisms,” so we must think that they also present recurrent features. But then, what distinguishes brokerage as a mechanism from brokerage as a process? The most appropriate way
to understand the distinction is to say that processes and mechanisms are identified in relation to an explanandum.
Typically, a particular episode of contention—the French Revolution, for instance—will
be considered as a large-scale process. This process, in turn, will be explained by reference to mechanisms, that is, to typical changes that have occurred within the temporal boundaries of the process, such as the mobilization of the Third Estate, the defection of the aristocratic elites, etc. Each of these mechanisms (e.g. elite defection, mobilization), in turn, can be considered as a new process to be explained. The process of mobilization of the Third Estate, for instance, can be explained by changes (mechanisms) that occurred within its temporal boundaries: identity shift among members of the Third Estate, brokerage by influential political entrepreneurs, etc.
At some point, explanations will also include reference to what MTT call “cognitive
mechanisms,” that is, alterations of individuals‟ perceptions and beliefs. But MTT do not pretend to offer such deep explanations and rather focus on what they call “relational mechanisms” that alter the connections among individuals and groups.
3. What’s wrong with DOC
During the years following its publication, DOC has been subjected to several criticisms. The most important are already present in a book symposium published in the journal Mobilization in
2003 with contributions from several major social movement scholars. Several points need to be highlighted here. Some commentators noticed that the “classical model” criticized by MTT is a straw man and does not capture the variety of approaches present in social movement research (Rucht 2003: 113). Others consider as arbitrary MTT‟s decision to limit their research to episodes of contention that involve state authority (Taylor 2003: 125).
More important for my point are criticisms directed at the causal-mechanistic framework. Although all commentators apparently accept the principles of mechanistic explanations, the same cannot be said of the particular accounts offered by MTT. Not only are MTT accused of crafting new concepts to refer to old mechanisms, but also of offering no principled way of identifying mechanisms. Ruud Koopmans (2003: 118), for instance, notes that “for each new chapter and paired comparison a new set of mechanisms is introduced, and formerly introduced ones rarely reappear.”
Two related features of MTT‟s work support the idea that recourse to mechanisms is ad
hoc. The first is the lack of methodical causal analysis to establish the relationship between a mechanism and a type of outcome. According to Koopmans (2003: 117), “most of the time, the authors do not show similar effects of one mechanism in widely different contexts, but different
effects of the same mechanism in different contexts.” Dieter Rucht (2003: 114) makes a similar point:
“While the authors demonstrate the existence of several of these mechanisms in each of the fifteen
episodes, they are less explicit and systematic in showing why some mechanisms are prevalent or
absent in specific episodes, how these mechanisms interact, and what consequences they produce
in which combination under which circumstances.”
In certain cases, the commentators stress that the explanatory mechanisms themselves call for an explanation: “Most of what MTT call mechanisms would be more fruitfully called processes. This is because they identify regular sequences of inputs and outputs (e.g. brokerage, identity formation) but rarely dig into how these worked.” (Oliver 2003: 120, see also Koopmans 2003: 117, Earl 2008: 357)
The second feature that seems to make the identification of mechanisms arbitrary is the extensive recourse to narratives. Although MTT argue that their objective is to go beyond “storytelling” (2001: 310), commentators note that “their book is filled with historical narratives” (Oliver 2003: 121) and that “the so-called mechanisms are not more than narrative aides”
(Koopmans 2003: 117). Although storytelling is not in itself objectionable, it apparently conflicts with the announced goal of proposing an explanatory framework that could be applied to a variety of situations. There is thus a tension in DOC that Pamela Oliver (2003: 121) describes in the following words: “A generalizing strategy focused on one process can never explain a historical episode, and should not try to. But conversely, a research program focused on explaining particular historical events cannot yield a general understanding of any process.”
(Oliver 2003: 121)
4. From mechanism sketch to mechanism schema?
Zenonas Norkus (2005) proposes a detailed analysis of the dilemma faced by MTT in their explanations of episodes of contention. His criticism draws on Machamer, Darden, and Craver‟s
distinction (2000) between mechanism sketches and mechanism schemata. They defined a
mechanism schema as “a truncated abstract description of a mechanism that can be filled with descriptions of known component parts and activities.” (Machamer et al. 2000: 15) By contrast, a
mechanism sketch is “is an abstraction for which bottom out entities and activities cannot (yet) be supplied or which contains gaps in its stages.” (Machamer et al. 2000: 18) Scientific progress, in
this view, is a movement from mechanism sketches to increasingly complete mechanism schemata. At the beginning, we have only a partial understanding of the different components of a mechanism, of their organization, or of the activities that they perform. Explanation is completed when all the components, activities, and organizational patterns that produce the explanandum phenomenon have been identified and accurately described.
According to Norkus, MTT fail to show that their mechanistic approach is capable of “schema instantiation,” that is, of “making a schema that is a mechanism sketch less abstract and applicable to a given case.” The reason is that they face what Norkus (2005: 368-369) calls the
“dilemma of specification”:
“…as one makes the attempt to elaborate Tilly‟s diagrammatic “mechanism sketches” into
descriptions of complete “mechanism schemata,” these sketches dissolve into many different
diagrams of lesser scope that are applicable only for some (or only one) episode from the initial set
of the episodes that instantiated the “mechanism sketch.” Not only the degree of abstraction, but
also the scope of the diagram, changes, so that it is not possible to consider them as depicting the
same mechanism or robust process.”
In other words, explanation is about filling the blanks of mechanism sketches, but doing this entails digging into the details of particular episodes, which are necessarily contingent:
“The detailed account of how the robust process of „actor constitution‟ took place, for example, in
the Civil Rights Movement in the U.S. in the 1950s, will show that the “micro-sequences” in the
particular episodes instantiating the putative mechanisms were contingent on some specific
circumstances that were not present in other episodes that were putatively considered as the
instantiations of the same mechanisms (for example, those during the Cultural Revolution in China
in 1966).” (Norkus 2005: 369)
The dilemma of specification explains why MTT are unable to offer mechanistic accounts that would be both complete and sufficiently general to apply to a wide variety of situations. It also
explains why accounts of particular episodes generally take the form of narratives:
“…more circumstantial research into the details of how some target outcome was produced
typically reveals that the process was a highly contingent or fragile one. The effort to transform a
mechanism sketch into a mechanism schema leads to process tracing, and the “mechanism” is no
longer distinguishable from a “mere” story. Evidently, therefore, mechanisms cannot supplant
stories if they do not work without stories or turn out to be stories in the “process tracing” used to
elaborate a mechanism sketch into a complete mechanism schema.”
I will argue below that the dilemma of specification must be taken seriously, but that it can be overcome—and has arguably been overcome in Tilly‟s later work—as long as explanandum
phenomena are well specified and realistic demands are put on what must be included in mechanism schemata.
5. From real mechanisms to heuristics
Reviewing the various criticisms on MTT, Chares Demetriou (2009) suggests another way of understanding their mechanistic approach. Demetriou recognizes the ambiguities present in MTT‟s work and the difficulties identified by the commentators, but he suggests that most of
these difficulties can be circumvented by introducing the distinction between explanatory
mechanism and real mechanism. In his view, explanatory mechanisms are not as strongly
committed ontologically as real mechanisms. The middle ground that he proposes is then to:
“…maintain theoretical commitment to mechanisms as ontologies and reduce the ambition of the
explanatory program through mechanisms, thereby treating the apprehension of real mechanisms
as an endeavor which, though elusive, orients and disciplines the delimitation of regularity. This