is network analysis stuck?

Here’s how I view the history of social network analysis:

  • Pre-history – Simmel (1900s) to Moreno (1930s): People start thinking about the “geometry” of social relationships.
  • Network science 1.0 – Harary, Heider, Freeman, etc. (1950s – 1970s): People learn to convert relational data into matrix algebra.
  • The holistic turn (1970s – 1980s): People start inventing measures of network structure (Bonacich, White).
  • Statistical theory of networks (1970s-2000s):  The creation of P* models, and later dynamic network models, to account for non-independence.
  • Socio-physics networks (2000s): Watts, Barbasi, and others from physics work on large scale properties of networks (e.g., power laws or small worlds).

So, by my account, the last major development in network analysis was about 10 years ago. Now, this isn’t to say that there isn’t excellent work, but it is normal science. Pick up a copy of Social Networks, or Network Science. You’ll see great articles, but they are usually investigating specific networks, or figuring out the details of some specific. Am I missing the next generation of network analysis? One possibility is that there will be new ideas coming from people doing experiments on networks for estimate causal effects. Other areas?

Adverts: From Black Power/Grad Skool Rulz

Written by fabiorojas

January 14, 2013 at 12:01 am

Posted in fabio, networks

10 Responses

Subscribe to comments with RSS.

  1. Well, that is one potted history of SNA.

    There is a whole other branch to SNA that is rooted largely in anthropology — J. Clyde Mitchell and the Manchester School of Africanists; Latin Americanists; Brits like Bott and Willmott; Americans like Nancy Howell Lee (Search for an Abortionist) and Charles Kadushin (ASR 1966) and so on — that has emphasized the content and substance of the network bonds more than their formal, graphical properties (even when that work is also statistical). From this angle, it is sometimes unclear what the value-added of the more formal modeling is.

    You can pick either branch as the more important one of SNA (that division drove grad students nuts when Ron Burt and I co-taught a SNA seminar sometime in the early ’80s), but SNA is both.


    Claude Fischer

    January 14, 2013 at 12:49 am

  2. @Claude: As a person who was weaned on Manchester school network analysis, I concur that I gave a very selective summary. My point wasn’t, though, to summarize all aspects of network analysis. Rather, I focused on its conceptual developments from a purely structural/technical view.

    You are clearly right in that I omitted substantive work. And I would say that the substantive side is doing great. All the time, I read great papers on how we’ve learned how network matter (or don’t) various contexts.

    But overall, if you were to teach a course on “basic tools for network analysis,” the last big development was probably the power law thing from a few years back. Is there any new idea or tool that substantive network analysis uses that less than ten years old? My only guess is p* or Sienna style models, but those are just extensions of p1.



    January 14, 2013 at 3:02 am

  3. Don’t forget the recent and renewed (mathematical) interest in bipartite networks (Opsahl & Panzarasa, 2009; Opsahl, Agneessens, & Skvoretz, 2010), though the basic conceptual ideas trace back to the 1940s (Davis et al., 1941).

    Moreover, some organization studies are moving away from the concept of social networks (i.e., networks of relations among individuals) towards an understanding of organizations as semantic networks (Oliver & Montgomery, 2008; Pentland & Feldman, 2007) or networks of communication episodes (Blaschke, Schoeneborn, & Seidl, 2012). In either one of these cases, however, the statistical analyses are pretty standard.

    * Blaschke, S., Schoeneborn, D., & Seidl, D. (2012). Organizations as Networks of Communication Episodes: Turning the Network Perspective Inside Out. Organization Studies, 33(7), 879–906.
    * Davis, A., Gardner, B. B., & Gardner, M. R. (1941). Deep South: A social anthropological study of caste and class. Chicago, IL: University of Chicago Press.
    * Oliver, A. L., & Montgomery, K. (2008). Using field-configuring events for sense-making: A cognitive
    network approach. Journal of Management Studies, 45(6), 1147–1167.
    * Opsahl, T., & Panzarasa, P. (2009). Clustering in weighted networks. Social Networks 31 (2), 155-163.
    * Opsahl, T., Agneessens, F., & Skvoretz, J. (2010). Node centrality in weighted networks: Generalizing degree and shortest paths. Social Networks 32 (3), 245-251
    * Pentland, B. T., & Feldman, M. S. (2007). Narrative networks: Patterns of technology and organization.
    Organization Science, 18(5), 781–795.



    January 14, 2013 at 8:08 am

  4. IMO, outside of epidemiology research is still insufficiently focused on the outcomes of network effects.



    January 14, 2013 at 2:51 pm

  5. I would argue that two major innovations are underway right now: Dynamic network analysis (i.e. R-SIENA) used to simulate the growth and change of social networks over time and more direct tests of influence/diffusion through Actor-Partner models (see Kreager and Haynie 2011). Those are both major statistical improvements that are expanding the theoretical relevance of the techniques, particularly in the social sciences.


    Nate Porter

    January 14, 2013 at 3:59 pm

  6. Well, you can’t really “pick up” an issue of Network Science, since the first one hasn’t come out yet, something that you probably know since some of the main editors are at Bloomington (although there’s at least one must-read article in the inaugural issue).

    If you look at that ToC, I think you will get a clue as to “what’s next”: dynamic, dynamic, dynamic + computational (which \neq “statistical”). See in particular here, and here. That’s why Nate is only half right: SIENA is only a tiny fraction of what is meant by “dynamic” in the current Network Science literature. The Oxford version of the Cambridge journal (, will no doubt feature much of the same work.



    January 14, 2013 at 5:26 pm

  7. My two cents: Over the last decade(ish) network analysis has moved from theoretical development towards testing if influence, contagion, and network processes do, in fact, operate in the real world as our theories predict. While SIENA and dynamic actor-network models are one direction I think the “hottest” area has been causal identification building form Manski’s seminal econometric work. I am thinking papers like the following:

    Sacerdote, B. 2001. “Peer Effects with Random Assignment: Results for Dartmouth Roommates.” The Quarterly Journal of Economics 116(2):681–704.

    Mouw, T. 2006. “Estimating the Causal Effect of Social Capital: A Review of Recent Research.” Annual Review of Sociology 32(1):79–102.

    Aral, S., L. Muchnik, and A. Sundararajan. 2009. “Distinguishing influence-based contagion from homophily-driven diffusion in dynamic networks.” Proceedings of the National Academy of Sciences 106:21544–49. Retrieved (

    Wang, D. J., and S. A. Soule. 2012. “Social Movement Organizational Collaboration: Networks of Learning and the Diffusion of Protest Tactics, 1960–1995 1.” The American Journal of Sociology 117(6):1674–1722.

    Each use sophisticated designs to make sure we are actually estimating network mechanisms. That said I do wonder if we are reaching a ceiling on clever causal identification papers.



    January 14, 2013 at 6:19 pm

  8. I think there are at least two major developments worth mentioning. First, the emergence of network economics in the ’00s, focusing on the game theoretical modeling of networks. Theoretically, also this movement seems to have lost its momentum, but the network experiments that you mention are in a sense the next step here, insofar as they test predictions from these models.
    Second, I think the latest new development is the emergence and analysis of huge online social networks. Just witness the ever growing number of online networks- and Twitter sessions at the Sunbelt conference.Substantively, this is somewhat different from the earlier socio-physics as it is very empirical and explicitly social, while it is accompanied by the influx of new people into the field, namely computer scientists such as Jon Kleinberg, Lada Adamic and Jure Leskovec.
    So no, I don’t think SNA is stuck, but maybe classical SNA as practiced by sociologists is at risk of being overrun by other fields.



    January 14, 2013 at 6:30 pm

  9. […] La sociologie des réseaux est-elle devenue une science normale? […]


  10. There are always innovations in social network theories. Honestly you’re neglecting all of the new data mining techniques involving social media. Hopefully, we’ll see some greater advances in the future.


    premed Thompson

    March 18, 2013 at 12:01 am

Comments are closed.

%d bloggers like this: