Peter Hedstrom and Richard Swedberg's Social Mechanisms: An Analytical Approach to Social Theory (1998) announced to the world the power of social mechanisms as a foundation for social explanations. It was based on a conference on this approach in Stockholm in 1996, and the volume includes contributions by outstanding authorities such as Thomas Schelling, Jon Elster, Aage Sorensen, and Arthur Stinchcombe (among others).
One person whom it does not include is Andrew Abbott. Abbott did indeed participate in the conference, but his contribution was not included in the volume when it appeared in 1998. The article did appear subsequently, however, in a special volume of Sociologica in 2007, with discussions by Delia Baldassarri, Gianluca Manzo, and Tommaso Vitale. (Here is a link to Abbott's essay and to his reply to discussants.) And the whole debate is worth reading. Abbott's piece is a serious challenge to the ontology of mechanisms. And each of the discussants in the Sociologica volume bring important criticisms and perspectives to the table.
Essentially Abbott takes issue with the mechanisms approach because it attempts to analyze the social world into fixed "atoms" of causal events (mechanisms and actors) rather than a more contextual ontology of relations and actions.
I shall consider the mechanisms movement from the viewpoint of a different theoretical tradition, one that focuses on the processual and relational character of social life and that traces its roots to pragmatism. (2)Abbott agrees with the CM approach in its rejection of a positivist search for statistical regularities among social characteristics -- what he refers to as the variables paradigm -- but he isn't persuaded by the ontology of mechanisms. He prefers actions to actors, and he prefers relationships linked to their contexts in time and place to portable mechanisms.
By the relational view I mean the notion that the meaning of an action is comprehensible only when it is situated in social time and place. A fundamental assumption of the mechanism view as set out here is that the meaning of a certain activity is given in itself. By contrast, the relational view assumes that the meaning of an action arises from its relations to other actions – both temporally, as a successor and a forerunner in coherent sequences of social events, and structurally, as a vertex in a synchronic ensemble of actions. Beneath this lies a more profound assumption that actions, not actors, are the primitives of the social process. The substratum of social life is interaction, not biological individuals who act. (7)
The most important part of Abbott's article is his positive argument for the primacy of a relational approach and actions-in-context instead of unitary actors. Much of this line of thought comes down to Abbott's view that actors and agency are deeply socially constructed; so it doesn't make sense to take the actor as a given who then deliberates about options.
Making interaction primitive makes it possible to give an account of the self. By making the self be continuously recreated in the flow of interaction we bring it out of the realm of assumptions and into that of investigation. At the same time, by making interaction primitive we allow for the endless interplay of cross-individual structural definitions of the flow of action, an interplay that is an evident fact in social life. (8)
For the relational account defines an act as a making of relations within a scene.... Social actors are cobbled together by actions that turn existing potential boundaries into actual ones. (9)
I am arguing the stronger point that the acting self is continuously remade in interaction and that the environment of possible endowments and contrasts--the environment of others and past experiences--provides the ground whence comes this remaking. (12)Another central thrust of Abbott's critique of the mechanism approach is that it is "reductionist" and depends on a rational-choice model of the actor. Essentially his line of argument is this: the mechanism approach requires microfoundations for macro-causal mechanisms; microfoundations presuppose rational actors; and therefore macro-facts are being reduced to facts about rational individuals. Each of these links is debatable, however, and in fact several of the discussants in the Sociologica volume question each of them. Essentially, many advocates of microfoundations (including me) insist on a richer theory of the actor than narrow economic rationality. And others -- for example, Dave Elder-Vass -- maintain that microfoundations don't imply reductionism either. (He prefers the idea of supervenience; I prefer explanatory autonomy; link.)
Delia Baldassarri highlights this issue in her comment, when she talks about the strengths and weaknesses of methodological individualism:
This approach has been quite successful in explaining "macro-phenomena that are emergent effects of the interdependent but uncoordinated actions of many individuals" [Mayntz 2004, 250]. The same approach has been less effective, however, in accounting for dynamics of identity construction, interest formation, boundary definition and institutional change, and in general, for social processes where macro-level states cannot be considered as given. (2)Gianluca Manzo's comment is the most extensive of the three in the Sociologica volume. His careful assessment leads him to conclude that Abbott overstates the incompatibility between "mechanisms" and "relations," and overstates as well the degree to which the analytical sociology perspective is "reductionist". Manzo advocates for what he calls "complex methodological individualism", and concludes that:
The "relational sociology" that Abbott defends is inconceivable without the "mechanismal sociology" (AS) he critiques.One feature of Abbott's article gives it its own particular relationality: he illustrates his meaning by describing a prolonged disagreement he had as a dean with the president of the University of Chicago over admissions recruitment strategies. His point, seemingly, is that the strategy chosen by the president presumes a rational-actor theory of college choice, and it further misunderstands the true priorities and considerations that motivate prospective UC students.
More important, we might phrase the difference by saying that the relational model is interested in how a student becomes a person who matriculates at the University of Chicago. (15)The inference: if UC is to continue to recruit the edgy, intellectual and quirky undergraduates it has been famous for, it will need to have a more nuanced understanding of the whole process. Or in the context of mechanisms versus relations: it won't do to think a marketing can pull the lever on a strategy (mechanism) and expect to get the desired result. Rather, we need a more contextualized and relational approach. There is no recipe of mechanisms -- grounded, moreover, in a rational expectations theory -- that will do the job. Instead, Abbott advises a relational, processual approach to college marketing: figure out the kinds of identities the desired segment of high school students are trying to make for themselves, and construct a series of experiences around that.
In many ways I find Abbott's essay more original than several in the original Hedstrom-Swedberg volume. So it's unfortunate that it wasn't included. It could have influenced the subsequent discussions very fruitfully.
Readers will also be interested in a recent collection edited by Pierre Demeulenaere that serves as a valuable follow-on to the original Hedstrom-Swedberg volume, Analytical Sociology and Social Mechanisms. It includes a piece by a German philosopher and sociologist whose work on this subject I admire, Michael Schmid, as well as contributions from Dan Sperber, Raymond Bouton, Jon Elster, Robert Sampson, and others. Schmid's Die Logik mechanismischer Erklarungen (2006) isn't available in English, which is unfortunate. The causal mechanisms approach has been picked up internationally, including an interesting book on the subject by the Italian sociologist Filippo Barbera (Meccanismi Sociali; Elementi di sociologia analitica; 2004).