Tracing the Productive
London School of Economics and Political Science
[**Original document page numbers preserved in brackets for citation purposes.**]
[Start Page 22]
In this paper, I conceptualize the notion of the productive continuity of a social mechanism. The continuity of a mechanism refers to the fact that each part of that mechanism should be causally connected to another; no part of the mechanism
is isolated. Moreover, each part of the mechanism should produce, i.e. cause, the next part1. The productive continuity of mechanisms forms an important part of philosophers of science Machamer, Darden, and Craver (MDC)’s “new mechanist”
conceptualization of biological mechanisms.2
Several process tracing methodologists have taken inspiration from the MDC conceptualization of mechanisms and the concept of productive continuity. Though I accept the notion of productive continuity as crucial for social mechanisms as it is for biological
mechanisms, I will show that the way one corroborates the productive continuity of social mechanisms is different from the way one corroborates it for biological mechanisms.
Philosophers of causal mechanisms disagree on what exactly productive continuity entails. I will consider two common answers, regularities and causal process
observations, before presenting my own, counterfactuals. While regularities, i.e. the recurrence of activities in other cases, are deemed necessary for biological mechanistic explanation by Machamer, Darden, and Craver,3 I show this
is not the case for social mechanistic explanation. Having thus discounted regularities as one source of evidence for the causal productivity of a social mechanism, I then turn to a common argument in the process tracing literature: that a social
mechanism can be supported by observing the implications of its components, i.e. causal process observations, in a particular case study4. I argue that this too is insufficient evidence for productive continuity. Finally, I show that counterfactual
evidence can provide evidence of the productive continuity of a social mechanism. I make the notion of counterfactual evidence more concrete by adopting the interventionist theory of causation set out by James Woodward.5
Philosophers of science Machamer, Darden, and Craver (herein “MDC”) characterize causal mechanisms6 as “entities and activities organized such that they are productive of regular changes from start or set-up to finish or termination conditions.”7 Entities
and activities are linked in productive continuity and the activities occur regularly, not just in one case, but in other cases too. I will discuss in detail what this regularity entails in section 3. For now, take the example of a biological causal
mechanism of chemical neurotransmission. In this mechanism, “a presynaptic neuron transmits a signal to a post-synaptic neuron by releasing neurotransmitter molecules that diffuse across the synaptic cleft, bind to receptors, and so depolarize the
post-synaptic cell.”8 Chemical neurotransmission’s entities include such things as the presynaptic and post-synaptic neuron. Entities, simply speaking, are things; they have a spatiotemporal location, and we can distinguish them in a straightforward
way from other entities. Entities are interdependent with the other part of a mechanism, the activities. The activities of the chemical neurotransmission mechanism
include diffusion and depolarization. The properties of the entities constrain what kind of activities they can undertake, and the kinds of activities undertaken constrain what kinds of properties the entities can have. MDC see activities as types of
causes (i.e. as ‘productive’); they “are producers of change; they are constitutive of the transformations that yield new states of affairs.”9
MDC argue that entities and activities are linked in what MDC call “productive continuity”: 10
"Complete descriptions of mechanisms exhibit productive continuity without gaps from the set up to termination conditions. Productive continuities are what make the connections between stages intelligible. If a mechanism is represented schematically by A
-> B -> C, then the continuity lies in the arrows and their explication is in terms of the activities that the arrows represent. A missing arrow, namely, the inability to specify an activity, leaves an explanatory gap in the productive continuity
of the mechanism."
Productive continuity is thus central to causal mechanistic explanation. If we cannot determine that one stage of the mechanism,
A, led to the next, B, we do not know whether
A is truly causally connected to B. However, as stated in the introduction, philosophers of causal mechanisms disagree on what exactly productive continuity entails, and this will concern me in what follows. I will now consider two common
regularities and causal
process observations, before presenting my own, counterfactuals.
for the productive continuity of social mechanisms
continuity supported by regularities
According to MDC, biological mechanisms only explain the link between a putative cause and effect of interest because the activities in these mechanisms are “regular,” i.e. because activities “work always or for the most part in the same way under the
same conditions.”11 In other words, the same activity working in some other biological system but in the same context will produce the same effect. The separate activities that together make up the mechanism of chemical neurotransmission
(e.g. diffusion, depolarization) will operate in the same way in like circumstances. Chemical neurotransmission is there- fore regular in the sense that it happens between many different sets of neurons in different biological systems to the same
However, not every new mechanist philosopher accepts MDC’s claim that regularity is necessary for productive continuity, and thereby for mechanistic explanation more generally. James Bogen suggests that although entities and activities are indeed organized
so that they are productive of changes, these changes do not need to be regular.12 After all, Bogen argues, there exist irregularly operating and stochastic activities and mechanisms, as well as activities and mechanisms that operate just
once. Though, in Bogen’s words, MDC believe that “generalizations describing natural regularities are essential components of causal explanations,”13 he argues that “causal productivity and regularity are by no means the same thing.”14
Another way of interpreting Bogen’s argument is as follows. Although the existence of regularities is one source of evidence for the causal productivity of a mechanism—since such regularities provide evidence that the link between some parts of the mechanism
is genuinely causal—this is not the only source of evidence for productive continuity. It would be misleading to think so, since excluding other sources of evidence means we cannot consider singly occurring activities as genuinely causal.
Let me now turn to why we ought to reject regularities as a key source of evidence for the productive continuity of
social mechanisms. First, consider MDC’s characterization of activities in terms of recurrence, i.e. as types of causes that describe something that acts “in the same way under the same conditions.”15 A presynaptic neuron’s release
of neurotransmitter molecules will take place in many such neurons in the body, and always in the same way, by virtue of the uniformity of such neurons.
However, the same demand for recurrence of the activity in other cases leads to difficulty in fields like political science. Arguably, there are no such causally homogeneous, stable entities and activities across a range of political science cases. Units
in social science mechanisms are not as clearly defined as biological entities; we cannot easily rely on a spatiotemporal location or set of boundaries to distinguish social “entities” and, similarly, social activities are not as straightforwardly
defined as are biological activities. There is, for instance, no reason to believe that a democratic government will always act in a similar way unless the same conditions are spelled out in excessive detail. There are more salient differences between
different democratic governments’ actions than there are be- tween the body’s different presynaptic neurons’ release of neurotransmitters.
In order to rely on regularities, one requires what Gerring has called “descriptive comparability,” the causal factors
X and effects of interest Y must “mean roughly the same thing across cases.”16 Finding descriptive comparability is particularly difficult to accomplish in the social sciences, since in social science we are dealing with aggregate,
social, non-individual concepts,17 what elsewhere has been described as “fuzzy” or “Ballung” concepts.18 Consider, for instance, the activity concept
making public a blue book, one of the key activities in the causal process that Kenneth Schultz claims linked the events behind Britain’s successful use of coercive
diplomacy during the Fashoda crisis.19 The blue book contained all the exchanges between Britain and France and communicated to the British public the uncompromising position of the government firsthand, which in turn made it difficult
for the government to renege on their position without incurring substantial political costs. Although one may speak of this activity more generally (there have been other instances in which a government made public their position in such a way),
there is no reason to assume that publishing a blue book hap- pens always or for the most part in the same way under the same conditions, nor is there reason to assume that it always or for the most part has the same effect. Thus, it is difficult
to see how a regularity, as defined by MDC, could be used to establish with certainty the causal productivity of Schultz’s purported causal process.
Note that this is not to say that general principles like regularities can never play a role in finding evidence for the productive continuity of a social mechanism. We may rely on more general political science studies to establish that, indeed, reneging
on promises made in a blue book incurs great political costs. However, such principles won’t take the form of MDC’s regularities, e.g., that making public a blue book “works always or for the most part in the same way under the same conditions.” I
will outline, in section 3.3.1 of this paper, the role of such general principles in providing evidence for productive continuity.
The second reason to reject regularities as necessary for social mechanistic explanation is as follows. In biology, we may see a straightforward relation between one mechanism and the process observed in a particular instance, e.g. the mechanism of protein
synthesis in general and the production of a protein in a particular cell under study. In the social sciences, things are less straightforward. Consider, for instance, Elisabeth Wood’s study of mobilization into the Frente Farabundo Martí para la
Liberación Nacional (FMLN) rebel forces during the Salvadoran Civil War.20 Wood argues that rural people mobilized because of three mechanisms: because they came to value participation per se, because of “defiance” (a refusal to acquiesce),
and because of “pleasure of agency” (the “positive affect associated with self-determination, autonomy, self-esteem, efficacy, and pride that come from the successful assertion of intention”21). Although each of the three mechanisms may
individually recur in other case studies, and thus have some degree of generality beyond the El Salvador study, the chain of events in El Salvador was a unique result; the same
process has not recurred in the same way in other conflicts. The causal mechanisms that Wood stipulates have interacted with the particular background conditions in El Salvador to produce this unique, single case and its entities’ activities.
To establish that the process in Wood’s study has productive continuity, i.e. that each part of the mechanism is causally related to (produces) the next, one cannot rely on the straightforward relation between one mechanism and one process as in biology.
To sum up, many of the processes traced in process tracing studies seemingly occur only once, and there is reason to believe that one can always find causally relevant differences between, for example, two conflicts, or one conflict at different points
in time. Moreover, MDC’s regularities often cannot sup- port the productive continuity of a social mechanism due to a lack of straightforward descriptive comparability between activities. As such, regularities as described by MDC are, at best, a possible
but not necessary source of evidence for productive continuity.
continuity supported by observable implications
Regularities, then, often cannot provide evidential support for the productive continuity of social mechanisms in the way MDC claim they do for biological mechanisms. Let us now consider a second source of evidential support for productive continuity
from the methodological literature.
Conceptualizing top-down process tracing, methodologists like Andrew Bennett and Jeffrey Checkel22 argue that one should first formulate a hypothesis about what may be the cause of an observed effect, and by what mechanisms cause and effect
are connected. One should subsequently try to pro- vide support for one’s own hypothesis, as well as refute any existing rival hypotheses, in a case study, using the observable implications of the hypothesized mechanisms. The observable implications
of the mechanism are generally called ‘causal process observations’ (CPOs) in the literature.23 We can think of CPOs as the salient observations a process tracer uses to evaluate a causal hypothesis. Bennett and Checkel define the observable
implications of mechanisms as “the facts and sequences within a case that should be true if each of the alternative hypothesized explanations of the case is true. Which actors should have known, said, and did what, and when? Who should have interacted
with, worried about, or allied with whom?”24
As an example, consider Kristin Bakke’s 2013 study of tactical innovation during the Chechen wars.25 Bakke hypothesizes transnational insurgents could have influenced local fighters through two mechanisms: relational and mediated diffusion.
To show that mediated diffusion and relational diffusion were indeed behind the radicalization of tactics in Chechnya, Bakke deduces several observable implications that should be true if these mechanisms were present. First, she shows the right background
conditions were present to make diffusion aimed at a radicalization of tactics possible (e.g. local fighters accepted the idea of direct attacks against civilians more over time); second, one of the most prominent hostage crises took place after training
camps and schools had been set up by transnationals and, as such, the chronology fits; third, Bakke presents some evidence that radical tactics were
being communicated to the local insurgents in these training camps and schools (e.g. evidence that hostage techniques were taught there and that videos were produced that taught suicide tactics).
The difficulty, however, with focusing on CPOs as evidence for process hypotheses, is that such evidence for the observable implications of mechanisms does not necessarily provide support for the
causal connection between the steps of the process.26 In other words, CPOs alone do not necessarily support the productive continuity of the mechanism. In Bakke, for instance, evidence for productive continuity is thin on the ground.
All that is required for Bakke, it seems, is that we observe the deductive implications of her hypothesized mechanisms in a case study. Yet it is less satisfying to state, for instance, that schools and training camps were built between the arrival
of transnational insurgents and the use of radical tactics, than it is to clearly link that it was in those schools and training camps that local insurgents became convinced that using, for example, suicide bombings is an effective and acceptable
In other words, if productive continuity is what one is looking for, CPOs are not the (best) way forward. CPOs do not commonly provide support for productive continuity, since they are not focused on providing support for the causal connection between
the links of a causal chain. In general, listing the observable implications of a mechanism’s entities and activities alone is not enough evidence to show that a putative cause and effect are indeed connected. Further evidence is necessary to show
that each event on the chain is causally connected to the events that directly precede and succeed it.
3.3 Productive continuity supported by
Now that we have seen regularities are unlikely to provide support for the productive continuity of social mechanisms, and that the use of causal process observations alone is insufficient, in the remainder of this paper I will consider a third source
of evidence process tracers may use to support productive continuity: counterfactuals, as per James Woodward’s interventionist theory of causation. Woodward has argued that any successful description of a cause-effect relationship must refer to causal
factors that can be manipulated to change the phenomenon under study.27 Specifically, X is a cause of Y if there exists some ‘intervention’ I that we can use to change X, so that X will then, in turn,
change Y without any interference of other factors linked to Y. In other words, using I, we can ascertain that
X made the change in Y happen. As I have shown in earlier work,28 Woodward’s theory implies that if we cannot specify an ‘intervention’ for each separate link of the chain of events, then these links are not genuinely causal,
and the process-tracing argument will fail. Evidence for interventions can support the productive continuity of social mechanisms, because it gives evidence for the arrows in a process A -> B -> C.
Woodward’s is a counterfactual theory of causation,29 and methodologists have previously considered what counter-factual evidence one may employ to support causal claims in the social sciences.30 A counterfactual is commonly defined
to be a claim of the form “If it had been the case that C (or not C), it would have been the case that E (or not E).”31 These frame- works, however, lack the concrete (Bayesian) evidential tests discussed by later process tracing methodologists.32 Wood-
ward’s theory on the other hand can be adapted to provide such a test.
In the Bakke study discussed above, the Woodwardian recommendation would be to go beyond the observable implications evidence (the CPOs) and show that e.g. if the local insurgents had not watched instructional videos on suicide tactics, then they
would not have used such tactics during later incidents. The intervention, here, is a technical way of conceptualizing the evidence one should provide to support this counterfactual.33 We would need to come up with an intervention that
prevents the locals from watching the videos, which in turn would prevent the use of such tactics. In particular, Bakke needs to ask, could we have prevented the local insurgents from watching suicide bombing videos, in a way that will not by itself
increase the use of radical tactics? (If we can only prevent the locals from watching suicide bombing videos by giving them a different source of information on such tactics, this arguably implies that they would have been led to using radical tactics
regardless of watching the videos.) Would they have used suicide bombings less if we had prevented them from watching such videos?
There are several kinds of interventions, according to Woodward’s theory: the actual (human) intervention, the natural intervention, and the theoretical intervention. In the particular context of Bakke’s work, a human intervention (akin to
an experiment in which the putative cause is actually intervened upon, e.g. in which we physically prevent the locals from watching the videos) is impossible, and more generally speaking, many process tracing case studies are unlikely to be compatible
with this technique. A natural intervention relies on finding a sufficiently similar case in which the cause was not present, to see what would happen to the effect. In Bakke’s case, one would need to find a (set of) conflict(s) that are similar in
every other way to the Second Chechen War, but where e.g. transnational insurgents are not present. For reasons mentioned above, it is unlikely that Bakke would find such a comparable case.
As such, the most likely type of intervention from Woodward’s framework that may provide us with a concrete (Bayesian) evidential test is the theoretical intervention: a process tracer would have to establish the counterfactual claim of what would
happen under an intervention if it were to be put in place. Woodward suggests that one way of doing so is through a hypothetical experiment. One would have to formulate an appropriate hypothetical experiment for the causal claim one is testing,
following the definition of an intervention variable I have described above. Then, one would collect data that tells us “what the results of the (…) hypothetical experiment would be if we were to perform the experiment, although in fact we don’t or
can’t actually perform the experiment.”34
As an indication of how one may go about such a hypothetical experiment, consider Bakke’s study again. She makes her causal claim regarding diffusion more salient by a counterfactual remark: “Suicide terrorism, in contrast [to hostage taking], does not
have a local historical template among the Chechens, despite centuries of conflict with central rulers. Thus in the absence of outside influence, it is unlikely that the Chechens would have turned to such a tactic.”35 But it is only a remark.
Establishing this counterfactual would be one way of corroborating what would happen under an intervention if it were to be put in place, i.e. what would happen if the Chechens had not been influenced by outside agents.
In order to establish this counterfactual, Bakke’s assumption must be that the Chechens before the arrival of the transnational insurgents are sufficiently similar to the Chechens after the arrival of the transnational insurgents, and thus that the diffusion
mechanisms that the transnational insurgents set in motion are the only cause of radicalization. Therefore, if there had been no transnational insurgents, nothing else would have caused the radicalization, and we would not have seen any use of suicide
Note that, in order to support such a detailed hypothetical intervention claim, and avoid black-boxing the links between watching videos and using suicide tactics,36 it is crucial that one is as detailed as possible. One must make explicit
“what it is about one part that links it in a causal sense to the next part.”37
Now let us briefly consider how a theoretical intervention may fit in with the evidential framework of later process tracing methodologists, particularly with the Van Evera terminology adopted by, amongst others, Beach and Pedersen, and Bennett and Checkel.38
Woodward’s framework for causation tells us that if an intervention for
the relation X -> Y can be shown to exist, X causes Y. This definition, in Woodward’s theory, means that if one finds an intervention I which answers to these criteria, it is unavoidable that X causes Y. In
theory, our subjective probability in that causal claim should be raised to 1. Finding an intervention is a sufficient but not necessary test that substantially raises our subjective probability for the causal claim. Failure to find an intervention
does not give us a reason to adjust our subjective probability for the causal claim one way or the other. As such, Woodward’s intervention fits perfectly with the smoking gun test.39
There are, however, limitations to the use of interventions as described above. Though in theory finding an intervention should raise our subjective probability in a process hypothesis to 1, in practice, finding an intervention will never actually confirm the
process hypothesis. We will raise our confidence in the process hypothesis more or less depending on the quality of our evidence for the intervention. How much our confidence is raised should be assessed on a case-by-case basis, and how one ought
to determine this is an important step for future work on process tracing methodologies. However, interventions remain a key source of evidence for the productive continuity of social mechanisms when compared to such sources as regularities and CPOs.
counterfactuals supported by regularities?
Before finishing, let me anticipate one potential objection to using counterfactual support for productive continuity. Authors like Fearon argue that the best support for counterfactuals comes through “invoking general principles, theories, laws, or regularities.”40 A
critic could thus suppose that getting counter factual evidence for productive continuity is no different from using regularities as per MDC’s framework. This would undermine my argument against regularities, since counterfactuals would then simply
be regularities in disguise.
The answer to this criticism lies in the way we interpret Fearon’s “regularities.” A straightforward regularity like the ones used in biology (e.g. that depolarization will work in the
same way in different places or cases under the same conditions) is inaccessible for singular social science cases. The support referred to by Fearon must therefore be of a different type, and indeed it is. Consider Fearon’s example of a general principle
one might invoke in describing the influence of nuclear weapons on the postwar world (from John Mueller’s 1988 study): “Wars are not begun out of casual caprice or idle fancy, but because one country or another decides that it can profit from (not
simply win) the war—the combination of risk, gain, and cost appears preferable to peace.”41 Comparing this general principle to the MDC regularity that presynaptic neurons, in the same circumstances, always release neurotransmitter molecules in a
certain way indicates the difference between support for the productive continuity of biological and social mechanisms. The former is exacting, describing a particular cause-effect relationship that recurs in the same way under the same conditions.
The latter is at best a guiding general principle, not a specific recurring relationship. As such, the demand for counterfactual support is not simply the call for regularities as per MDC in disguise. Indeed, following my logic in the preceding section,
I would argue that the use of general principles as in Mueller’s study are only supportive of productive continuity if such general principles can be used to sup- port a specific intervention claim that implies that if the cause had not been present,
the effect would not have followed, all other things being equal.
In this paper, I have analysed three kinds of support for assessing the productive continuity of social mechanisms: regularities, causal process observations, and interventions. I have shown that regularities, because of the dissimilarities between social
and biological activities, often cannot provide support for social mechanisms’ productive continuity. I have also shown that causal process observations do not commonly provide support for productive continuity, since they are not focused on providing
support for the causal connection between the links of a causal chain. Finally, I outlined how counterfactual evidence can provide support for social mechanisms, in the form of Woodward’s interventions, and finished with some concrete requirements
that this evidence must meet. Future work should determine exactly how Woodwardian interventions can provide a smoking gun test for process tracing hypotheses.42
Bakke, Kristin M. 2013. “Copying and Learning from Outsiders? Assessing Diffusion from Transnational Insurgents in the Chechen Wars.” In Transnational Dynamics of Civil War, edited by Jeffrey
T. Checkel. Cambridge: Cambridge University Press, 31–62. Beach, Derek and Rasmus B. Pedersen. 2013. Process-Tracing Methods: Foundations and Guidelines. Ann Arbor: University of Michigan Press.
Beach, Derek. 2016. “What are we actually tracing? Process tracing and the benefits of conceptualizing causal mechanisms as systems.” Qualitative and Multi-Method Research: Newsletter of the
American Political Science Association’s QMMR Section vol. 14, no. 1-2.
Bennett, Andrew and Jeffrey T. Checkel. 2015. “Process Tracing: from Philosophical Roots to Best Practices.” In Process
Tracing: from Metaphor to Analytic Tool, edited by Andrew Bennett and Jeffrey T. Checkel. Cambridge: Cambridge University Press, 3–38.
Bogen, James. 2005. “Regularities and Causality; Generalizations and Causal Explanations.” Studies in History
and Philosophy of Biological and Biomedical Sciences vol. 36, no. 2: 397–420.
Bogen, James. 2008. “Causally Productive Activities.”
Studies in History and Philosophy of Science vol. 39, no. 1: 112–123.
Brady, Henry E. and David Collier. 2010.
Rethinking Social Inquiry: Diverse Tools, Shared Standards. 2nd ed. Lanham: Rowman & Littlefield.
Cartwright, Nancy and Norman M. Bradburn. 2011. “A Theory of Measurement.” In The Importance of Common Metrics
for Ad- vancing Social Science Theory and Research: Proceedings of the National
Research Council Committee on Common Metrics. 53– 70.
Cartwright, Nancy and Rosa W. Runhardt. 2014. “Measurement.” In Philosophy of Social Science: a New
Introduction, edited by Nancy Cartwright and Eleonora Montuschi. Oxford: Oxford University Press, 265–287.
Darden, Lindley. 2008. “Thinking Again about Biological Mechanisms.” Philosophy of Science vol. 75, no. 5: 958–969.
Fearon, James D. 1991. “Counterfactuals and Hypothesis Testing in Political Science.” World Politics vol. 43, no. 2: 169–195.
Gerring, John. 2005. “Causation: A Unified Framework for the Social Sciences.” Journal of Theoretical
Politics vol. 17, no. 2: 163–198.
Kincaid, Harold. 2009. “Causation in the Social Sciences.” In The Oxford Handbook of Causation, edited by Helen Beebee, Christo- pher Hitchcock and Peter Menzies. Oxford: Oxford University Press. 726–743.
Machamer, Peter 2004. “Activities and Causation: the Metaphysics and Epistemology of Mechanisms.” International
Studies in the Philosophy of Science vol. 18, no. 1: 27–39.
Machamer, Peter, Lindley Darden, and Carl F. Craver. 2000. “Thinking about Mechanisms.” Philosophy of Science vol. 67, no. 1: 1–25. Mueller, John. 1988. “The Essential Irrelevance of Nuclear Weapons: Stability in the Postwar World.” International Security vol.13:
Psillos, Stathis. 2004. “A Glimpse of the Secret Connexion: Harmonizing Mechanisms with Counterfactuals.” Perspectives
on Science vol. 12, no. 3: 288–319.
Runhardt, Rosa W. 2015. “Evidence for Causal Mechanisms in Social Science: Recommendations from Woodward’s Manipulability Theory of Causation.” Philosophy of Science vol. 82, no. 5: 1296– 1307.
Schultz, Kenneth A. 2001. Democracy
and Coercive Diplomacy. Cambridge: Cambridge University Press.
Tetlock, Philip E. and Aaron Belkin. 1996.
Counterfactual Thought Experiments in World Politics: Logical,
Methodological, and Psychological Perspectives. Princeton: Princeton University Press.
Van Evera, Stephen. 1997. Guide
to Methods for Students of Political Science. Ithaca: Cornell University Press.
Wood, Elisabeth. 2003. Insurgent
Collective Action and Civil War in El Salvador. New York:
Cambridge University Press.
James. 2002. “What Is a Mechanism? A Counterfactual Account.” Philosophy of Science vol. 69, no. S3: S366–S377.
Woodward, James. 2003. Making
Things Happen: a Theory of Causal Explanation. Oxford: Oxford University Press.
Woodward, James. 2015. “Methodology, Ontology, and Interventionism” Synthese vol. 192: 3577–3599.
1 Though productive continuity is a property of causal mechanisms themselves, one can also demand the productive continuity of a description of a causal mechanism between some putative cause and effect of interest. If a description of a mechanism
is unclear about how certain parts of the mechanism are linked to the rest, that description does not satisfactorily explain the relation between the cause and effect under study.
2 Machamer, Darden, and Craver 2000.
3 cf. Bogen 2005, 2008, Machamer 2004, Woodward 2002.
4 The type of process tracing I am concerned with here fits best with what is called theory-testing process tracing (Beach and Pedersen 2013), in which one traces a causal mechanism through case-specific observable implications of this mechanism’s
top-down process tracing (Bennett and Checkel 2015), in which one tests type- level causal hypotheses about the mechanisms producing a particular “process” with case study data.
5 Woodward 2003.
6 Note that MDC specialize in philosophy of biology, particularly molecular biology and neurobiology. MDC express the suspicion however that their analysis “is applicable to many other sciences, and maybe even to cognitive or social mechanisms” (Machamer,
Darden, and Craver 2000, 2). My conclusions about social mechanisms further on in this paper will show the scope of MDC’s theory is more limited than they suspect; MDC’s new mechanist philosophy may not reach far beyond biology.
7 Machamer, Darden, and Craver 2000, 3. Several process tracing methodologists have taken inspiration from the MDC conceptualization of mechanisms. Beach and Pedersen, for instance, write that “[e]ach of the parts of the causal mechanism can be conceptualized
as composed of entities that undertake activities” (Beach and Pedersen 2013, 29).
8 Machamer, Darden, and Craver 2000, 3.
9 Darden 2008, 962.
10 Machamer, Darden, and Craver 2000, 3.
11 Machamer, Darden, and Craver 2000, 3.
12 Bogen 2005 and 2008.
13 Bogen 2005, 397.
14 Bogen 2005, 397.
15 Machamer, Darden, and Craver 2000, 3.
16 Gerring 2005, 184.
17 cf. Kincaid 2009.
18 Cartwright and Bradburn 2011, Cartwright and Runhardt 2014.
19 Schultz 2001.
20 Wood 2003.
21 Wood 2003, 235.
22 Bennett and Checkel 2015.
23 See Brady and Collier 2010.
24 Bennett and Checkel 2015, 30.
25 Bakke 2013.
26 cf. Runhardt 2015.
27 Woodward 2003.
28 Runhardt 2015.
29 My demand for further counterfactual evidence for the links of the causal chain fits with Stathis Psillos’ distinction between mecha- nistic and counterfactual causation, and his belief that mechanistic causal claims must rely on counterfactual causal
claims (Psillos 2004).
30 cf. Fearon 1991, Tetlock & Belkin 1996.
31 Fearon 1991, 169.
32 cf. Beach and Pedersen 2013; Bennett and Checkel 2015.
33 In Woodward’s framework, I is an intervention for this causal relation if, firstly, I causes viewing of videos of suicide bombings; secondly, I acts as a switch for the local insurgents’ increased use of suicide bombings (i.e.,
makes whether the insurgents use this tactic independent of any other influences); thirdly,
I does not directly or indirectly cause the increased use of suicide bombings itself; fourthly, I is statistically independent of any factor A not on the path I Z
Y that links the viewing of videos of suicide bombings, Z, to the in- creased use of suicide bombings, Y. One should show that all four requirements are met by a (theoretical) intervention if it is to be accepted as evidence for a process
hypothesis. For more details, see Runhardt (2015). (These strict requirements put on an intervention also imply that even in cases of redundancy, we can assess whether the putative cause was indeed causally relevant. These cases are caught under the
second requirement, which fixes other influences on the effect of interest. As such, being explicit about the requirements one puts on an intervention responds to Beach’s (2016) objection that counterfactual accounts break down in cases of redundancy.)
34 Woodward 2015, 3587.
35 Bakke 2013, 58.
36 Beach’s (2016) criticism, in this symposium, is that counterfactual understandings of mechanisms black-box or mask causal logics.
37 See Beach (2016) in this symposium.
38 Van Evera 1997, Beach and Pedersen 2013, Bennett and Checkel 2015.
39 According to Van Evera, we can use the predictions a claim implies (e.g., the predicted existence of an intervention) to test whether that claim explains a causal relation satisfactorily. A smoking gun test is a set of “predictions of high uniqueness
and no certitude” (Van Evera 1997, 40). Passing a smoking gun test strongly corroborates a claim (due to its uniqueness), but failing to pass such a test does not strongly undermine a claim. After all, as Van Evera says, “suspects not seen with a
smoking gun are not proven innocent.” (Van Evera 1997, 40). Finding evidence for an intervention can strongly corroborate a causal claim.
40 Fearon 1991: 176.
41 Fearon takes this example from Mueller’s 1988 study. Mueller 1988: 68-69, as cited by Fearon 1991: 184.
42 This work should further highlight that the level of detail required in interventionist evidence avoids black-boxing or masking the logics of a causal process. See footnote 36.