Andy Abbott QOTD: Cumulativity and Progress in Sociology

Andrew Abbott is one of the rock stars of contemporary sociology. His book, The System of Professions was agenda-setting for the study of professions, and his later work on thinking about puzzles, mechanisms, time and history makes a regular appearance on methods syllabi at all levels. Institutionally, Abbott has long served as editor of one of the two most influential journals in the field, the American Journal of Sociology. Given his institutional positions and topics of research, Abbott has some incredibly insightful things to say about sociology as a discipline and a profession. I had the pleasure of reading two short pieces by Abbott this weekend.

The first, discussed here by Daniel Little, is Abbott’s criticism of the mechanisms movement (for lack of a better name) spearheaded by Hedstrom and Swedberg. Abbott argues against mechanisms and for relational approaches (although he sees both as improvements on the old general linear reality model he criticized eloquently in the 1980s). Abbott argues that the basic unit of social life is actions (and interactions) not individuals (or entities). This view dovetails nicely with my own, ANT-inflected, sensibilities, and I recommend the article and Daniel Little’s explication of it. But a second piece, stumbled across while searching for the first, really caught my eye: Reconceptualizing Knowledge Accumulation in Sociology.

In this article, Abbott asks a deceptively straightforward but incredibly important question: what would it mean to say that sociology is a cumulative science?

What is real cumulativity as opposed to mere repetition? Is cumulativity a matter of piling up facts? of developing theories? of paradigm shifts? What are the alternatives to cumulativity as models of scientific life? (57)

Abbott argues that in the last 100 years, sociology’s access to data has grown tremendously, while our basic frameworks have remained more or less similar. So, what have we accumulated? Abbott begins by showing that the age-structure of citations in top sociology journals has actually moved against expectations: on average, we cite older articles than we used to (e.g. in 1977, 60% of citations in AJS were to pieces less than 10 years old, in 2005 it was only 45%). So, we appear not to be so myopically focused on the recent past, as some of the narrative on cumulation might lead us to expect. But, as Abbott notes, it’s really not clear how to interpret these measures, except perhaps that the post-war quantitative turn makes older articles more intelligible now than those from the pre-war period (as the methods are more similar to those still en vogue).

Abbott moves from this quantitative analysis to some speculations about a 25 year lifespan for research programs/paradigms in sociology, which he connects in the end to generations of scholarship and the apparent progress we observe. In other words, really interesting ideas spawn successful careers and follow-up projects that expand the scope of the original ideas and solve pesky theoretical problems, but peter out by the end of that generation of scholarship, such that we see progress on the work we began as young scholars but by the time we finish our careers the field has moved on (rather than continuing work on the same issues under the same headings). From here, Abbott moves to his most interesting discussion: “Cumulativity as a Theoretical Issue”. Abbott argues that cumulativity involves some sort of commensurability but also some sort of abstract theoretical framework, not just tables of interpretable datapoints. And here we finally arrive at the rather long QOTD:

But the project of abstraction clearly has intrinsic limits, because large parts of social life concern things that are specifically set up not to be law governed; the game of human life is after all to assign new meanings to old things. That’s how we differ from ants. As a result, each time we social scientists think we have found a law of human behavior, we can suddenly see all the ways in which people create meaning by violating it. It is the same as when a bunch of people have cut a shortcut across the university quad, and once it becomes normative and we cave in and pave it, people begin to take shortcuts from the middle of it across another green space and so on. The limit is simple; you pave the quad. The green space–which in my analogy is the interesting stuff about human behavior– simply disappears. Similarly, the project of finding the ultimate law-like basis to social behavior–the project pursued by many scholars, (my friend the late Roger Gould, for example)–is a logical impossibility. It is a dream that leads us on and on and that eventually will simply lead us back to where we started and then run off laughing. (63)

Abbott’s beautiful prose is a variation on the problem of Asimov’s psychohistorians: people understand the rules of the game they are playing and actively seek to subvert them. Knowledge about the game is part of the game. We do things because we’re told not to; we resignify old ideas and things simply because we have been told what they are supposed to mean. Or, to put a performativity spin on things, social scientific models of the world reshapes the world in ways that may improve or worsen the fit between the model and the world. But Abbott also sets up with this quote a delightful concluding passage, after a segue into some discussions about the sort of computational model that might best describe sociology:

From the point of view of my computational models of scholarship, of course, what we really are is indeed little pieces of something much larger, something that may not be going anywhere at all. … In the neural net view, we are somewhat random processors taking in an idiosyncratic array of inputs and doing a fairly idiosyncratic array of things with them, with only a vague idea of nature of the attractor that defines our part of the social scientific inquiry space and a sneaking suspicion that it, too, changes more rapidly than we think. Either way, to take ourselves for what we really are is pretty alienating, in terms of the traditional ideologies of science with which we grew up; we are not heroes of discovery, but random explorers in Borges’ endless, labyrinthine library. So perhaps an ideology of cumulation is how we make sense of this to ourselves, finding a level of the system with whose direction and pace we feel ourselves comfortable. We can then feel that at least for a while we are part of something that is going somewhere. Maybe that’s the best we can hope for. (65)

I, for one, am quite content to explore the endless library* hoping, but uncertain, that we might find something worth reading in the next room over.

In any event, this short (10p) article should be required reading for graduate sociology methods courses. The vocabulary and set of questions it provides would do a lot to help students with different perspectives on “sociology as social science” to communicate with each other and determine where their disagreements originate.

* It’s worth noting that for Borges, this fate would not be perhaps quite as tragic as for others: “Yo, que me figuraba el Paraíso / Bajo la especie de una biblioteca.”



  1. Good post Dan. Abbott is always interesting.

    I came to graduate school in sociology because I was interested in understanding how to improve public education. More generally, I believe sociologists can improve decision-making through accumulated knowledge about social processes and methods which can help answer many specific questions. Building a cumulative science is difficult, but it is possible, and I think it *should* be one of our goals. I’m open to considering other goals as well, but I don’t really want to share a discipline with someone who isn’t trying to contribute to a cumulative science.

    • Thanks for the comment!

      Two thoughts:
      First, Abbott distinguishes improvement from cumulation:
      “It is also important to notice that cumulation is not the same thing as improvement. Indeed, improvement and cumulation are often opposed. For we can know more about something while forgetting particular details we used to know, and we can create vastly improved methodologies that despite their overall preferability cannot do particular things our older methodologies could and that therefore negate all the knowledge our older methodologies produced.” Not really addressing your point that well, but perhaps of interest.

      Second, the question isn’t just whether or not some of us want to contribute to a cumulative science and some of us want to go off into the woods and howl at the moon while sacrificing quantitative datasets to the dark Foucauldian gods.* The question is, what does it mean to do cumulative social science and how do we know if we’ve gotten there? Do our perceptions of progress in our own research and that of our allies give us a good indication? What is to be made of the fact that we seem to reinvent concepts, or close cousins, with frequency? Is this a positive working things out or a degenerative trend? And how much hope should we have of getting generalized knowledge good for “all times and places”? If not that, then what? Those are the questions Abbott is getting at and that I think practicing sociologists have a lot of disagreement about. I tend to fall on the side of those who think that knowledge is really hard, always partial, and rarely applicable as widely as we’d like. To the extent we know much about how the world works in general it is because there are systems and technologies in place that force the world to look similar in different places (and those systems, as the world-society folks will tell us, are largely similar in name but not in practice). Does that make me anti-“cumulative science”? Or just pessimistic?

      * Though that sounds like a fun weekend activity if anyone’s up for it!

      • Dan, I think you raise a lot of good points. Cumulative science is hard. No doubt certain facts, phenomena, and methodologies are forgotten by people over the years.

        I think the idea of a cumulative science is related to the idea of rigor. I would like to see sociologists making more testable predictions. Ideally they would be stored in a public database. One of the goals of such a project would be to discover what sociologists actually believe. The other would be to create incentives to be correct. Obviously only a minority of sociologists would participate, but I think we’d learn a lot, and that this predictive knowledge would be more cumulative than most sociology.