orgtheory.net

the failure of the network approach or how structuralism returned to where harrison white started

The publication of Harrison White’s notes in Sociologica is one of the coolest things I’ve ever seen in an academic journal! The notes themselves are fascinating, although if you’ve read some of Identity and Control and Markets from Networks they’re not entirely new. Perhaps more fascinating though are the commentary around the notes. Marco Santoro offers one of the most impressive and succinct summaries of the history of network analysis that I’ve seen (by focusing on White’s early contribution), and Michael Schwartz’s reflections on the class are really interesting.

What’s remarkable about White’s notes, in retrospect, is how complete his view of networks and categories were for their time. That’s not to say that White had it all figured out way back in the early 1960s, but all of the necessary parts were there. The “elementary types of social structure” – networks, categories/identities, frames – are all present in these notes. It’s also remarkable that it took so long for structural sociology to reincorporate all of these elements after a long period of networks-dominated structural theory.

Both from Santoro’s and Schwartz’s account we learn that the real motivator behind White’s Social Relations 10 class was the dominant Parsonian theory that focused so heavily on “attributes and attitudes.” I had no idea that White’s theorizing was such a strong reaction to Parsons’s structural-functionalism. I realized that Goffman and the symbolic interactionists were very critical of Parsons’s approach, but Schwartz makes it clear that White was equally resistant and even hostile (perhaps this is why I see complementarity between White’s theory and symbolic interactionism).

In those days, he was not even a little bit diplomatic about this disdain. I can remember (in a graduate course) him savaging Parsons while Chad Gordon (a Parsonsian and his co-teacher) squirmed around trying to figure out a defense. For us as graduate students, that disdain was the hookfor everything else. We could easily see the failures of the “attributes and attitudes” approach in the various texts we were reading (by ourselves or with his help), and this made us hungry for an alternate view that could better explain the evidence.

The rejection of Parsons united a group of grad students around White. In White’s “catnets” approach they saw a new kind of sociology emerging that could change the way sociologists think about individuals and their relation to society. The basic idea was that individuals could be understood as mutually embedded in personal relationships (i.e., networks or “nets”) and categorical domains (i.e, identities or “cats”). The mutual reinforcement of these two constraints created the basic framework that shaped sociability. White’s approach was revolutionary in the sense that it avoided functionalist theorizing while offering an alternative empirical methodology to attitudinal research.

Students of networks and culture will note that not all of White’s ideas were picked up by the cadre of grad students that came out of those “heady” times. Rather quickly the project quickly coalesced around the idea that structuralism should be the empirical analysis of networks. Santoro notes that many of his students became almost solely methodologically focused, developing the tools that would come to be the backbone of network analysis in the years to come. Even White became preoccupied block modeling for a time. The cultural component of White’s original “catnets” thesis was largely left behind (although see Breiger’s duality thesis as a nice step in that direction). In fact, one might say that the “catnets” idea wasn’t picked up again until the ecologists revamped some of his ideas with new terminology – see especially the work of Miller McPherson, whose “ecology of affiliation” approach was the beginning, I think, of an attempt to reinvigorate the “duality of networks and identities” idea.

Why did White’s grad students get so caught up in the relational component of catnets and lose the cultural component? Who knows really. One explanation is that the students wanted to get as far away from “attributes and attitudes” as they could and network materiality was a nice bunker in which to build reinforcement. Michael Schwartz’s experience illustrates:

Another of Harrison’s great innovations was “problem sets,” just like in Math courses. One problem set, I remember, led to a fine illustration of the hub structure of networks. The students were required to choose their two best friends and have those two choose their two best friends and etc., going out, I think, to six removes. Most of the networks folded back to the student tracing them, but one of my students’ had a network which consisted of something like 20 people stretched out with almost no reciprocal ties, and zero people choosing my student. I took one look at the network she had traced and felt a great sorrow come over me, as I realized that she was a very lonely, unloved person, without any real group of friends. And then I was won over to the fundamental theory of network analysis, that structure expresses and/or creates content.

For at least the next twenty years, the network folks would attempt to show how all kinds of “content” were endogenous to network structure. There is still a lot of that research still being done (especially in organizational studies), much of it very well-done and interesting. But while the network approach generated significant empirical insights, it never seemed to live up to its revolutionary potential. Recently, some have worried that the network approach was in the doldrums. One might speculate that network exhaustion creeped in because the entire project departed from its theoretical foundations too early. White’s original intent, I think, was to show that network structure and identities were mutually constitutive. You couldn’t understand how one evolved without the other. Nets without cats wasn’t a theory at all.

More recently the catnets concept has been rejuvenated, some of it by White’s own work in the books mentioned above. But others, like Paul DiMaggio, John Mohr, Ann Mische, have done their best to reunite culture and networks. A 2000 issue of Poetics pushed the idea even further, even publishing a paper by White on modeling discourse. Ron Breiger organized another special issue of Poetics in 2005 that focused more exclusively on “cats” – how classification affects markets. Thus, it appears that 40 years later, structural sociology is finally embracing the totality of catnet, which perhaps explains why it took so long for the original notes to be published.

Written by brayden king

June 5, 2008 at 5:11 pm

17 Responses

Subscribe to comments with RSS.

  1. Very well said Brayden. I think your analysis nails the basic issue (or problems) right on the head.

    And yes, the publication of those Catnet notes is pretty darn cool!

    Like

    Omar

    June 5, 2008 at 5:22 pm

  2. Excellent post! About why “cats” were left behind: I’d guess that the desire for normal science drew people toward the elements of the theory that were most susceptible to formal models. Note that the cohort coming out of the early days produced highly math-accesible concepts: block models, centrality, brokerage, generalized exchange, etc. There are all easily expressible as properties of adjacency matrices, once you get the basic network idea down. In contrast, getting beyond “Catcs” as discrete value variables is not obvious. I’d probably give more credit to the cognitive anthropologists, and their kin in soc, who have sophisticated theories of group membership. Either way, the success of “catnets” (“netcats?”) is remarkable.

    Like

    fabiorojas

    June 5, 2008 at 5:23 pm

  3. I think that’s about right Fabio. One of the insiders (Paul DiMaggio) pretty much laid that problem bare in an excellent 1993 book chapter entitled “Nadel’s Paradox Revisited.”

    Also, I’m looking forward to John Martin’s presentation at this year’s theory miniconference (which, not coincidentally, Ann Mische decided to dedicated to the issue of methods and formal representation): “Don’t Bogart that Joint Homomorphic Reduction, Or, When You Turn Social Life into a Bunch of Ones and Zeros, From Where Do You Get the Ones?” (?!) Which from what I gather from the funky title I assume it’s going to deal with similar issues.

    Like

    Omar

    June 5, 2008 at 5:31 pm

  4. Why did White’s grad students get so caught up in the relational component of catnets and lose the cultural component?

    Formal tractability, combined with rhetorical opposition to the cultural/categorical/normative excesses of their opponents.

    The fragmentation of sociological theory in the late 60s and early 70s can be seen as a series of efforts to escape the Parsonian synthesis. One of the early books to put that fragmentation in context, by Ian Craib, makes this case. Even though he was unaware of the network structuralists, his general argument applies to them too — viz, that many of these reactions (from ethnomethodology to conflict theory to interactionism to structuralism) could be seen as breaking away by focusing heavily on a part of the Parsonian synthesis and emphasizing that to the exclusion of much else. This made them fail in the same kind of way, as their partiality became a limitation and an inhibition to further development after a point. The difficulty was that even if the Parsonian view was sterile, it was comprehensive.

    I just received my copy of the new edition of Identity and Control in the post, and what’s interesting to see is how it, too, now tries to present a comprehensive vocabulary for theorizing about social formations, explicitly including the various cultural aspects. White also insists that no part of the theory can really work in the long term without the others, even if they might need to be digested one piece or cut at a time to make them comprehensible. The modeling aspect is very much in the background and the emphasis is on teaching you the concepts and vocabulary.

    Like

    Kieran

    June 5, 2008 at 5:56 pm

  5. That Santoro piece is very useful.

    Like

    Kieran

    June 5, 2008 at 6:02 pm

  6. “4 out of 5 orgtheory guys agree that formal tractability resulted in an overemphasis on the structural potential of network configurations”

    Teppo, buddy, you with us here?

    Like

    fabiorojas

    June 5, 2008 at 6:04 pm

  7. Teppo is traveling to Denmark and may not be able to respond immediately. So I’ll respond for him. Yes, he agrees with us. He is also ready to announce a formalization of an individualist theory that encapsulates all of White and all of Parsons in a single theorem.

    Like

    brayden

    June 5, 2008 at 6:23 pm

  8. I’m looking forward to John Martin’s presentation at this year’s theory miniconference (which, not coincidentally, Ann Mische decided to dedicated to the issue of methods and formal representation): “Don’t Bogart that Joint Homomorphic Reduction, Or, When You Turn Social Life into a Bunch of Ones and Zeros, From Where Do You Get the Ones?” (?!) Which from what I gather from the funky title I assume it’s going to deal with similar issues.

    This problem reliably gets returned to every twenty years or so in network theory, and also in the analysis of stratification and mobility. And this has been the case at least since Nadel’s formulation of the problem, as Paul’s article points out. The core issue is always the same: it is hard to do a formal analysis of the overall structure of a system of positions, and harder still to do an analysis of the flow of persons or resources with such a system. But both of these problems remain much more tractable than a similar analysis of the origin and evolution of the system of positions itself, and because it is so much harder we are constantly tempted to fall back to the first two approaches, even though the third is theoretically fundamental and — inescapably — entails questions about the meaning of the structure and its positions to those who constitute it.

    Like

    Kieran

    June 5, 2008 at 6:26 pm

  9. Kieran, to which Craib book are you referring?

    Like

    brayden

    June 5, 2008 at 6:39 pm

  10. “He [Teppo] is also ready to announce a formalization of an individualist theory that encapsulates all of White and all of Parsons in a single theorem.”

    That’s why he’s in charge.

    Like

    fabiorojas

    June 5, 2008 at 6:52 pm

  11. Fabio: Sorry — I am in Copenhagen for the week, will have to catch up on all this when I get back…momentarily buried.

    Like

    tf

    June 5, 2008 at 7:11 pm

  12. Too late, TF, Brayden has already signed you up for the new post-Parsons Harrison White hegemony!! By the time you get back, we’ll be working through the counter movement to White. Next week, we’ll be so advanced that we’ll upload our collective brains into the blog and emerge as the new group mind of sociology, and Scatter will be a mere comment in an infinite post describing our most pure sociological thoughts.

    Like

    fabiorojas

    June 5, 2008 at 7:31 pm

  13. This is a fascinating post!

    ( disclaimer: my undergrad adviser was Scott Boorman )

    I’m very much in agreement with Fabio that “the desire for normal science” is largely the reason for “catnets” only becoming expressed in limited forms. But I think the remark of Fabio’s should be explored much more thoroughly.

    In computer science, where I live, there’s been a Harrison White-like movement for years that’s tried to put category theory at the center of programming. In math, there’s always been debate about category theory’s value. Its founders, famously, in a move of hopeful vaccinating self-deprecation, called it “abstract nonsense”. Category theory began in the US at the same time as Bourbaki, and then was picked up by Grothendieck, the leader of the generation after Bourbaki, with topos theory.

    The usual attacks against category theory and functional programming are either that they’re vapid fancy expressions of what people already know, or they’re opaque and hard to read or evaluate, i.e., violate usability and clear style. Both attacks can be viewed as attacks from a point of view of “normal science”. The attacks become stronger when viewed as an attempt to create an elite at a time when computer science starves for new students. What’s of course shocking or unsurprising given the public backdrop is that corps like Microsoft have grabbed functional programming researchers with heavy math background in large quantity.

    I’m not sure how the attacks work out in sociology. But someone like Mark Granovetter has probably lived his whole career positioning himself in relation to the attacks. It’s not a pretty story to date. IMHO “catnets” is still waiting for a Grothendieck.

    Like

    Tony

    June 8, 2008 at 1:34 pm

  14. Oops, my blogging dyslexia has appeared again. The sentence above, “the attacks become stronger”, should be “the attacks against formal methods, which include category theory and functional programming, become stronger when formal methods are viewed as an attempt to create an elite while computer science starves for new students”. One can easily replace “formal methods” with “catnets”, and “computer science” with “sociology”, in the above, as anyone with a categorical frame of mind would do almost immediately.

    Much of the mindset of category theory is about finding structural connections between more or less independently elaborated constructions, which connections implicate each in the other. To which a normal response from a specialist in such constructions is “it’s all greek to me”. If such a specialist is a publication firewall, guess the outcome.

    Like

    Tony

    June 8, 2008 at 1:42 pm

  15. “IMHO “catnets” is still waiting for a Grothendieck.”

    Indeed, I think Harrison White might be the sole legitimate claimant to the role that Grothendieck (still?) has in math. Right now, though, his big ideas are in texts that most people find insanely hard to read. People sit around and praise them, but I bet few have read them.

    The genius of Grothendieck was to apply his ideas in wide range of sitiations that most mathematicians cared about. And to help people really understand the ideas. That’s the key point. Though White has written a lot, most people know him from block modelling, Identity and Control, and recently the Networks/Markets book. Each of these is an achievement, but so far, there’s been no attempt at retaill dissemination. At best, they know a block model, and highly regard the books, but can’t use them for their own purposes.

    Like

    fabiorojas

    June 8, 2008 at 2:16 pm

  16. i agree.

    Like

    Tony

    June 8, 2008 at 4:56 pm

  17. […] In American Sociology, this general rule is probably most applicable to network analysis.  The basic innovations (and innovators) of the so-called “Harvard-School” centered around Harrison White were certainly part of an informal network endowed with their own set of, never published, shadow texts in which the basic programmatic theses were written (For a nice discussion of this see Santoro 2008, and this post and this post). […]

    Like


Comments are closed.

%d bloggers like this: