Complex Contagions and the Diffusion of Innovations: Evidence from a Small-N Study

Abstract

The recent literature on “complex contagions” challenges Granovetter’s classic hypothesis on the strength of weak ties and argues that, when the actors’ choice requires reinforcement from several sources, it is the structure of strong ties that really matters to sustain rapid and wide diffusion. The paper contributes to this debate by reporting on a small-N study that relies on a unique combination of ethnographic data, social network analysis, and computational models. In particular, we investigate two rural populations of Indian and Kenyan potters who have to decide whether to adopt new, objectively more efficient and economically more attractive, technical/stylistic options. Qualitative field data show that religious sub-communities within the Indian and Kenyan populations exhibit markedly different diffusion rates and speed over the last thirty years. To account for these differences, we first analyze empirically observed kinship networks and advice networks, and, then, we recreate the actual aggregate diffusion curves through a series of empirically calibrated agent-based simulations. Combining the two methods, we show that, while single exposure through heterophilious weak ties were sufficient to initiate the diffusion process, large bridges made of strong ties can in fact lead to faster or slower diffusion depending on the type of signals circulating in the network. We conclude that, even in presence of “complex contagions,” dense local ties cannot be regarded as a sufficient condition for faster diffusion.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9

Notes

  1. 1.

    A synthetic measure of reachability suggested by the literature on small-world topologies referred to in the introduction would have been the average path length, i.e., the average number of ties (computed over the shortest paths) one must go through to reach two nodes (Watts 2004). However, not all the networks that we compare here are fully connected (see, in particular, Figs. 3 and 4), thus making the average path length a less informative measure.

  2. 2.

    For an “objective” comparison of firing techniques, which is in fact in line with our potters’ own perceptions, see Gosselain (1992) and Livingstone (2001).

  3. 3.

    Unless differently specified, all network visualizations and statistics were produced using the igraph R package (version 1.0.1).

  4. 4.

    It should be noted in passing that a few previous empirical studies (see, for instance, Rao et al. 1980) found that social fragmentation (in terms of sub-castes and clans) among Indian Hindu farmers slowed down diffusion of agricultural innovations within this community.

  5. 5.

    During the interview, MAD acknowledged that the approach she followed to manufacture the new shape makes her pots more fragile.

  6. 6.

    In particular, in the Indian social context, it was reported that in Sar (one case), Sangasni (five cases), and Bipalsur (one case), some sons had had to push their fathers to adopt the kiln; in Tanawra (one case), the son of one potter was only able to adopt the kiln once he had left his father’s household and established his own pottery business. This suggests that the authority on the final decisions about whether or not to change technology may reflect the age pyramid.

  7. 7.

    Let us clarify that, as we model the diffusion of technological innovation with economic consequences, one may want to explicitly represent a cost-benefit analysis at the agent-level (for an example of this, see, for instance, Young 2006, 2011). As we have explained in section “Analytical Framework and Network Measures,” however, there are good empirical reasons to think that, in both social contexts, there is no by-group systematic difference with respect to the evaluations of the payoffs of the innovations under scrutiny. For this reason, we treat the kiln’s cost-benefit ratio as constant across the two communities and omit it from the micro-level decision mechanisms we postulate.

  8. 8.

    This is a form of robustness check in the sense that, in the absence of empirical information to adjudicate between different modeling choices, we assess the extent to which the core mechanisms of our models are stable across different adopter selection criteria (for a short introduction to robustness analysis, see Railsback and Grimm 2011:302–6).

  9. 9.

    Again, since we do not have precise and extensive empirical information on how often potters exchange information and help about the kiln, we decided to run the simulation under different interaction rates. This is a form of sensitivity analysis in the sense that we assess the extent to which the core mechanisms of our models are sensitive to the value of a numerical parameter for which empirical setup is not possible (for a short introduction to sensitivity analysis, see Railsback and Grimm 2011:293–297).

  10. 10.

    The 192 * 4 (averaged) simulated curves generated by our models cannot be shown for lack of space. Needless to say, they are available upon request. From time to time, we will discursively rely on these simulated diffusion curves to justify the selection of our preferred models.

  11. 11.

    Let us stress that this model variant is a very good fit with the observed diffusion curves under an interaction rate of 1. This means that the simulation postulates that, over the entire temporal window under observation, i.e., 26 years, about 9360 interactions (i.e., about once a day) included discussions about the kiln. Given the size of the Muslim family network, the geographic area involved, and the family habits of the Muslim community, where family members meet very frequently, this seems a realistic flow of kiln-related interactions (compared to the interaction rate of other model variants showing similar aggregate curve fitness).

  12. 12.

    The differences actually concern the interaction rate—which turned out to be lower for the simulated Hindu population—and the selection criterion for the adopter(s) from whom information flows at each iteration—which turned out to be centrality, rather than expertise. As to the interaction rate, given the smaller size of the Hindu family network, the geographical distances between spouses and their original village, and the family habits of the Hindu community, which, as we have noted, tend to limit meeting opportunities, a lower flux of kiln-related interactions—namely 2340 (i.e., about 90 interactions a year)—seems a realistic difference compared to the simulated Muslim population. As to the selection criterion for the source of influence, it is interesting that expertise works more poorly than centrality in the simulation for Hindu potters. We noted in section “Analytical Framework and Network Measures” that it is likely that the best potters were not necessarily the most deeply involved in the diffusion process.

References

  1. Abramson, C. M., Joslyn, J., Rendle, K. A., Garrett, S. B., & Dohan, D. (2017). The promises of computational ethnography: improving transparency, replicability, and validity for realist approaches to ethnographic analysis. Ethnography, 19(2), 254–284. https://doi.org/10.1177/1466138117725340.

    Article  Google Scholar 

  2. Aral, S., & Van Alstyne, M. (2011). The diversity-bandwidth trade-off. American Journal of Sociology, 117(1), 90–171.

    Article  Google Scholar 

  3. Baldassarri, D. (2015). Cooperative networks: altruism, group solidarity, reciprocity, and sanctioning in Ugandan farmer organizations. American Journal of Sociology, 121(2), 355–395.

    Article  Google Scholar 

  4. Banerjee, A., Chandrasekhar, A. G., Duflo, E., & Jackson, M. O. (2013). The diffusion of micro-finance. Science, 341(6144), 1236498. https://doi.org/10.1126/science.1236498.

    Article  Google Scholar 

  5. Barabási, A.L. & Bonabeau, E. (2003) Scale-Free Networks. Scientific American, 288, 60–69.

    Article  Google Scholar 

  6. Bittles, A. H., & Hussain, R. (2000). An analysis of consanguineous marriage in the Muslim population of India at regional and state levels. Annals of Human Biology, 27(2), 163–171.

    Article  Google Scholar 

  7. Böhm, T., & Hillmann, H. (2015). A closed elite? Bristol’s Society of Merchant Venturers and the abolition of slave trading. Political Power & Social Theory, 29, 147–175.

    Article  Google Scholar 

  8. Borgatti, S. P., Everett, M. G., & Johnson, J. C. (2013). Analyzing social networks. London: SAGE Publications Limited.

    Google Scholar 

  9. Boudon, R. (1986). Theories of social change: a critical appraisal. Berkeley and Los Angeles: University of California Press.

    Google Scholar 

  10. Boudon, R. (2003). Beyond rational choice theory. Annual Review of Sociology, 29(1), 1–21.

    Article  Google Scholar 

  11. Boudon, R. (2014). What is context? Kölner Zeitschrift für Soziologie und Sozialpsychologie, 66(Suppl), 17–45.

    Article  Google Scholar 

  12. Bruch, E., & Atwell, J. (2015). Agent-based models in empirical social research. Sociological Methods & Research, 44(2), 186–221.

    Article  Google Scholar 

  13. Centola, D. (2010). The spread of behavior in an online social network experiment. Science, 329(5996), 1194–1197.

    Article  Google Scholar 

  14. Centola, D. (2011). An experimental study of homophily in the adoption of health behavior. Science, 334(6060), 1269–1272.

    Article  Google Scholar 

  15. Centola, D. (2015). The social origins of networks and diffusion. American Journal of Sociology, 120(5), 1295–1338.

    Article  Google Scholar 

  16. Centola, D. (2018). How behavior spreads: the science of complex contagion. Princeton: Princeton University Press.

    Google Scholar 

  17. Centola, D., & Macy, M. W. (2007). Complex contagions and the weakness of long ties. American Journal of Sociology, 113(3), 702–734.

    Article  Google Scholar 

  18. Clarke, D. L. (1968). Analytical archaeology. London: Methuen & Co ltd..

    Google Scholar 

  19. Coleman, J. S., Katz, E., & Menzel, H. (1957). The diffusion of an innovation among physicians. Sociometry, 20(4), 253–270.

    Article  Google Scholar 

  20. Davis, D. D. (1983). Investigating the diffusion of stylistic innovations. In M.B. Schiffer (Ed.), Advances in archaeological method and theory (Vol. 6, pp. 53–89). Academic Press. https://doi.org/10.1016/B978-0-12-003106-1.50007-9

    Google Scholar 

  21. Deffuant, G., Huet, S., & Amblard, F. (2005). An individual-based model of innovation diffusion mixing social value and individual benefit. American Journal of Sociology, 110(4), 1041–1069.

    Article  Google Scholar 

  22. Delre, S. A., Jager, W., Bijmolt, T. H. A., & Janssen, M. A. (2010). Will it spread or not? The effects of social influences and network topology on innovation diffusion. Journal of Product Innovation Management, 27(2), 267–282.

    Article  Google Scholar 

  23. DiMaggio, P., & Garip, F. (2012). Network effects and social inequality. Annual Review of Sociology, 38(1), 93–118.

    Article  Google Scholar 

  24. Erikson, E. (2017). Social networks and macrosocial change. Annual Review of Sociology, 43(1), 229–248.

    Article  Google Scholar 

  25. Ermakoff, I. (2017). On the frontiers of rational choice. In C. Benzecry, M. Krause, & I. A. Reed (Eds.), Social theory now (pp. 162–200). Chicago: The University of Chicago Press.

    Google Scholar 

  26. Fink, C., Schmidt, A., Barash, V., Kelly, J., Cameron, C., & Macy, M. (2016). Investigating the observability of complex contagion in empirical social networks. In International AAAI Conference on Web and Social Media. https://www.aaai.org/ocs/index.php/ICWSM/ICWSM16/paper/view/13143. Accessed 18 Dec 2017.

  27. Ganguli, K. K. (1983). Cultural history of Rajasthan. Delhi: Sundeep Prakashan.

    Google Scholar 

  28. Gosselain, O. P. (1992). Bonfire of the enquiries. Pottery firing temperatures in archaeology: what for? Journal of Archaeological Science, 19(3), 243–259.

    Article  Google Scholar 

  29. Gosselain, O. P. (2016). To hell with ethnoarchaeology. Archaeological Dialogues, 23(2), 215–228.

    Article  Google Scholar 

  30. Granovetter, M. (1973). The strength of weak ties. American Journal of Sociology, 78(6), 1360–1380.

    Article  Google Scholar 

  31. Hägerstrand, T. (1971). A Monte Carlo approach to diffusion. European Journal of Sociology, 6(1), 43–67.

    Article  Google Scholar 

  32. Hamberg, K., & Gargiulo, F. (2014). Virtual fieldwork: modeling observer bias in kinship and marriage networks. Journal of Artificial Societies and Social Simulation, 17(3), 2.

    Article  Google Scholar 

  33. Hamberg, K., Houseman, M., & White, D. R. (2011). Kinship network analysis. In J. Scott & P. J. Carrington (Eds.), The Sage handbook of social network analysis (pp. 533–549). London: Sage.

    Google Scholar 

  34. Hanneman, R. A., & Riddle, M. (2011). A brief introduction to analyzing social network data. In J. Scott & P. J. Carrington (Eds.), The Sage handbook of social network analysis (pp. 331–339). London: Sage.

    Google Scholar 

  35. Hollstein, B. (2011). Qualitative approaches. In J. Scott & P. J. Carrington (Eds.), The Sage handbook of social network analysis (pp. 404–416). London: Sage.

    Google Scholar 

  36. Kalter, F., & Kroneberg, C. (2014). Between mechanism talk and mechanism cult: new emphases in explanatory sociology and empirical research. Kölner Zeitschrift für Soziologie und Sozialpsychologie, 66(Suppl 1), 91–115.

    Article  Google Scholar 

  37. Keuschnigg, M., Lovsjö, N., & Hedström, P. (2017). Analytical sociology and computational social science. Journal of Computational Social Science. https://doi.org/10.1007/s42001-017-0006-5.

    Article  Google Scholar 

  38. Kramer, C. (1997). Pottery in Rajasthan. Ethnoarchaeology of two Indian cities. Washington and London: Smithsonian Institution Press.

    Google Scholar 

  39. Krings, G., Calabrese, F., Ratti, C., & Blondel, V. (2009). Urban gravity: a model for inter-city telecommunication flows. Journal of Statistical Mechanics: Theory and Experiment, https://doi.org/10.1088/1742-5468/2009/07/L07003.

    Article  Google Scholar 

  40. Lambiotte, R., Blondel, V. D., de Kerchove, C., Huens, E., Prieur, C., et al. (2008). Geographical dispersal of mobile communication networks. Physica A: Statistical Mechanics and its Applications, 387(21), 5317–5325.

    Article  Google Scholar 

  41. Lazega, E., & Pattison, P. E. (1999). Multiplexity, generalized exchange and cooperation in organizations: a case study. Social Networks, 21(1), 67–90.

    Article  Google Scholar 

  42. Léon-Medina, F. J. (2017). Analytical sociology and agent-based modeling: is generative sufficiency sufficient? Sociological Theory, 35(3), 157–178.

    Article  Google Scholar 

  43. Livingstone, S. A. (2001). Bonfire II: the return of pottery firing temperatures. Journal of Archaeological Science, 28(9), 991–1003.

    Article  Google Scholar 

  44. Manzo, G. (2014a). The potential and limitations of agent-based simulation: an introduction. Revue Française de Sociologie, 55(4), 653–688.

    Article  Google Scholar 

  45. Manzo, G. (2014b). Data, generative models, and mechanisms: more on the principles of analytical sociology. In G. Manzo (Ed.), Analytical sociology: actions and networks (pp. 4–52). Chichester: Wiley.

    Google Scholar 

  46. Miller, D. (1985). Artefacts as categories. A study of ceramic variability in Central India. Cambridge: Cambridge University Press.

    Google Scholar 

  47. Mitschele, A. (2014). Identity and opportunity in early modern politics: how job vacancies induced witch persecutions in Scotland, 1563–1736. In G. Manzo (Ed.), Analytical sociology: actions and networks (pp. 151–169). Chichester: Wiley.

    Google Scholar 

  48. Opp, K.-D. (1999). Contending conceptions of the theory of rational action. Journal of Theoretical Politics, 11(2), 171–202.

    Article  Google Scholar 

  49. Padgett, J. F., & Ansell, C. K. (1993). Robust action and the rise of the Medici, 1400-1434. American Journal of Sociology, 98(6), 1259–1319.

    Article  Google Scholar 

  50. Palloni, A. (2001). Diffusion in sociological analysis. In J. B. Casterline (Ed.), Diffusion processes and fertility transition: selected perspectives (pp. 66–115). Washington, DC: National Academy Press.

    Google Scholar 

  51. Railsback, S. F., & Grimm, V. (2011). Agent-based and individual-based modeling: a practical introduction. Princeton: Princeton University Press.

    Google Scholar 

  52. Rao, G. A., Rogers, E. M., & Singh, S. N. (1980). Interpersonal relations in the diffusion of an innovation in two Indian villages. Indian Journal of Extension of Education, 16(1–2), 19–24.

    Google Scholar 

  53. Rivera, M. T., Soderstrom, S. B., & Uzzi, B. (2010). Dynamics of dyads in social networks: assortative, relational, and proximity mechanisms. Annual Review of Sociology, 36(1), 91–115.

    Article  Google Scholar 

  54. Robins, G. (2011). Exponential random graph models for social networks. In J. Scott & P. J. Carrington (Eds.), The Sage handbook of social network analysis (pp. 484–501). London: Sage.

    Google Scholar 

  55. Robins, G. (2015). Doing social network research: network-based research design for social scientists. London: Sage.

    Google Scholar 

  56. Rogers, E. M. (2003). Diffusion of innovations. New York: Free Press.

    Google Scholar 

  57. Rossman, G., Chiu, M. M., & Mol, J. M. (2008). Modeling diffusion of multiple innovations via multilevel diffusion curves: payola in pop music radio. Sociological Methodology, 38(1), 201–230.

    Article  Google Scholar 

  58. Roth, C., Gargiulo, F., Bringé, A., & Hamberg, K. (2013). Random alliance networks. Social Networks, 35(3), 394–405.

    Article  Google Scholar 

  59. Roux, V. (2012). Spreading of innovative technical traits and cumulative technical evolution: continuity or discontinuity? Journal of Archeological Method and Theory, 20(2), 312–330.

    Article  Google Scholar 

  60. Roux, V. (2015). Standardization of ceramic assemblages: transmission mechanisms and diffusion of morpho-functional traits across social boundaries. Journal of Anthropological Archaeology, 40, 1–9.

    Article  Google Scholar 

  61. Roux, V. (2017). Not to throw the baby out with the bathwater. A response to Gosselain’s article. Archaeological Dialogues, 24(2), 225–229.

    Article  Google Scholar 

  62. Roux, V., Bril, B., & Karasik, A. (2018). Weak ties and expertise: crossing technological boundaries. Journal of Archaeological Method and Theory, 25(4). https://doi.org/10.1007/s10816-018-9397-8.

  63. Rowe, W. L. (1960). The marriage network and structural change in a north Indian community. Southwestern Journal of Anthropology, 16(3), 299–311.

    Article  Google Scholar 

  64. Ryan, B., & Gross, N. C. (1943). The diffusion of hybrid corn in two Iowa communities. Rural Sociology, 8(1), 15–24.

    Google Scholar 

  65. Saraswati, B. (1979). Pottery-making cultures and Indian civilization. New Delhi: Abhinav Publications.

    Google Scholar 

  66. Smith, J. A., & Burow, J. (2018). Using ego network data to inform agent-based models of diffusion. Sociological Methods & Research, 1–46. https://doi.org/10.1177/0049124118769100.

  67. Snijders, T. A. B. (2011). Network dynamics. In J. Scott & P. J. Carrington (Eds.), The Sage handbook of social network analysis (pp. 501–514). London: Sage.

    Google Scholar 

  68. Stovel, K., & Shaw, L. (2012). Brokerage. Annual Review of Sociology, 38(1), 139–158.

    Article  Google Scholar 

  69. Strang, D., & Soule, S. A. (1998). Diffusion in organizations and social movements: from hybrid corn to poison pills. Annual Review of Sociology, 24(1), 265–290.

    Article  Google Scholar 

  70. Tubaro, P., & Casilli, A. A. (2010). “An ethnographic seduction”: how qualitative research and agent-based models can benefit each other. Bulletin of Sociological Methodology/Bulletin de Méthodologie Sociologique, 106(1), 59–74.

    Article  Google Scholar 

  71. Valente, T. W. (1999). Network models of the diffusion of innovations. New York: Hampton Press.

    Google Scholar 

  72. Valente, T. W., & Davis, R. L. (1999). Accelerating the diffusion of innovations using opinion leaders. The ANNALS of the American Academy of Political and Social Science, 566(1), 55–67.

    Article  Google Scholar 

  73. Van den Bulte, C., & Lilien, G. L. (2010). Medical innovation revisited: social contagion versus marketing effort. American Journal of Sociology, 106(5), 1409–1435.

    Google Scholar 

  74. van Duijn, M. A. J., & Huisman, M. (2011). Statistical models for ties and actors. In J. Scott & P. J. Carrington (Eds.), The Sage handbook of social network analysis (pp. 459–484). London: Sage.

    Google Scholar 

  75. Watts, D. J. (1999). Networks, dynamics, and the small-world phenomenon. American Journal of Sociology, 105(2), 493–527.

    Article  Google Scholar 

  76. Watts, D. J. (2004). The new science of networks. Annual Review of Sociology, 30(1), 243–270.

    Article  Google Scholar 

  77. Watts, D. J., & Doods, P. S. (2007). Influentials, networks, and public opinion formation. Journal of Consumer Research, 34(4), 441–458.

    Article  Google Scholar 

  78. Watts, C., & Gilbert, N. (2014). Simulating innovations. London: Edward Elgar Publishing.

    Google Scholar 

  79. Watts, D. J., & Strogatz, S. H. (1998). Collective dynamics of ‘small-world’ networks. Nature, 393(6684), 440–442.

    Article  Google Scholar 

  80. Wejnert, B. (2002). Integrating models of diffusion of innovations: a conceptual framework. Annual Review of Sociology, 28(1), 297–326.

    Article  Google Scholar 

  81. White, D. R., & Houseman, M. (2003). The navigability of strong ties: small worlds, tie strength, and network topology. Complexity, 8(1), 72–81.

    Article  Google Scholar 

  82. White, D. R., & Johansen, U. (2006). Network analysis and ethnographic problems: process models of a Turkish nomad clan. Lanham: Lexington Books.

    Google Scholar 

  83. Young, P. (2006). The diffusion of innovations in social networks. In L. E. Blume & S. N. Durlauf (Eds.), The economy as a complex evolving system (Vol. III). Oxford: Oxford University Press.

    Google Scholar 

  84. Young, P. (2009). Innovation diffusion in heterogeneous population: contagion, social influence, and social learning. American Economic Review, 99(5), 1899–1924.

    Article  Google Scholar 

  85. Young, P. (2011). The dynamics of social innovation. Proceedings of the National Academy of Sciences, 108(4), 21285–21291.

    Article  Google Scholar 

Download references

Acknowledgements

In India, the support of the Rupayan Sansthan was invaluable. We are grateful to Kuldeep Kothari for resolving all the logistical problems as well as Meet Kaur Gulati, Anil Sharma, Ira Sisodia, and Lakshman Diwakar for their assistance in the field. In Kenya, we are grateful to Jacqueline Kawira for her assistance in the field. Last but not the least, we would also like to thank all the Indian and Kenyan potters for their availability and their unfailing kindness. Mohamed Cherkaoui, Ivan Ermakoff, Cyril Jayet, and Jörg Stolz read earlier versions of the manuscript and provided instructive written remarks. We also thank JAMT referees for their invaluable critical feedback. We are grateful to Peter Hamilton for his careful linguistic revision. Usual disclaimers apply.

Funding

This work was supported by the ANR (The French National Agency for Research) within the framework of the program CULT (Metamorphosis of societies—“Emergence and evolution of cultures and cultural phenomena”), project DIFFCERAM (Dynamics of spreading of ceramic techniques and style: actualist comparative data and agent-based modeling) (no. ANR-12-CULT-0001-01).

Author information

Affiliations

Authors

Contributions

The article arises from a common theoretical and methodological effort. Authors were however differently involved in the specific tasks needed to reach the final product. In particular, [Freda Nkirote M'Mbogori] especially contributed to field investigations in Kenya; [Valentine Roux] especially contributed to research design and field investigations in India; [Simone Gabbriellini] especially contributed to qualitative data coding, modeling and simulation, and data analysis; [Gianluca Manzo] especially contributed to research and method design, analysis of the literature, modeling, data analysis, and article writing.

Corresponding author

Correspondence to Gianluca Manzo.

Appendices

Appendix 1. Ethnographic appendix

Appendix 1.1

In the Jodhpur region, we visited 18 rural villages and 1 urban center (Jodhpur). Some of these villages are inhabited only by Hindu potters (Bisalpur, Jodhpur, Sathin, Pachpadra); some others only by Muslim potters (Sar, Salawas, Rudakali, Tanawra, Sangasni, Ramasni, Rohat, Dhava, Banjara, Boranada Basni); and still others are mixed (Siwana, Mokalsar, Banar, Palasni). The villages’ sizes range from those where only one potter household is present (as in Bipalsur) to larger villages with 40 active potters (such as Banjara, Ramasni, or Pachpadra)—on average, 14 potter households were active in the villages under scrutiny. Among mixed villages, the Muslim/Hindu ratio also differs: in Banar, Muslim potters are in the overwhelming majority (14 versus 1); in Mokalsar, the reverse holds (3 versus 10); Siwana and Palasni present a much more balanced ratio (2[1] Muslims versus 3[2] Hindus, respectively).

Appendix 1.2. Excerpts of interviews with Indian potter

(From interview with BIK in Sar) In 1987, BIK, Talu’s son, had a brother-in-law who used to work in Ahmedabad and who informed him about possible well-paid seasonal work. So, he went to Ahmedabad and worked for a few months in workshops run by Hindu potters. He discovered how to use the kiln there. When he came back, he decided to build one of them because it was less effort and more rapid than the open firings. (From interview with Fakhir in Salawas) FAK was the first one to adopt the kiln in 1987. He saw the kiln when he went to Sar to meet relatives. BIK had just built it. (From interview with BIE in Sangasni) The adoption of the kiln started in Sangasni in 1990 with two potters, BIE and HAN, who had relatives in Salawas. After discussing with CHA, brother-in-law of HAN and cousin of BIE, CHA came to Sangasni to help them to build the kiln. (From interview with RAM in Banar) In 1990, RAM, son of NUR, 20 years old, went to Ahmedabad to sell pots. Over there, he saw the kiln and the Hindu potters explained to him how it worked. However, he did not adopt it even though he could see advantages. In 1996, he went to Salawas for the wedding of the youngest sister of his father. During this social event, he met CHA who agreed to come to Banar to help him build the kiln. (From interview with GAN in Ramasani) The kiln was adopted in 1995 by four households at a time (SAD, MOH, UZI, and GAN). They went to Salawas for a family event. They saw the kiln and found that it was much faster than open firing. They asked IDU from Salawas to come to Ramasani to make the kilns. IDU came, made the kilns, and showed them how to make the firing. He made a demonstration attended by dozen potters.

Appendix 1.3. Marriage rules

Both the Muslim and the Hindu potters are endogamous communities, patrilineal and virilocal (see Saraswati 1979; Ganguli 1983). They practice village exogamy, with the brides usually leaving their natal communities to join their grooms’ households. As a general rule, these brides originate from families which do not have a patrilineal link with the groom’s family—the families the daughter can, and cannot, marry with are called genait and bhaipa, respectively. Despite these common rules, significant differences exist in the marriage rules between the two communities. As for the Muslim potters, they descend, in each village, from a common, well-identified ancestor, who was usually born in one of the neighboring villages—for instance, Boranada Basni is the village ancestor of Salawas (11 km away) which is the village ancestor of Sar (10 km away). Muslim potters preferentially marry maternal first cross-cousins, a general rule in the Indian-Muslim communities (Bittles and Hussain 2000). This preferential marriage rule goes hand in hand with the exchange of women between genait villages, which enables Muslim potters to maintain symmetrical alliances and avoid differences of status between wife-givers and wife-takers. Hindu potters are embedded in a different social structure and have different marriage rules. They belong to three main sub-castes, the Banda, the Purubiya, and the Maru (the latter being present nowadays only in Jodhpur). These three sub-castes are endogamous, implying that there cannot be kinship ties between Purubiya and Banda (and Maru). Each of these sub-castes is in turn subdivided into gotras, or clans, which are exogamous. Several sub-castes can inhabit the same place, but, as a general rule, marriage alliances are made in different villages and sometimes with remote villages where only one alliance will occur (for instance, the brothers of a same family will marry women from different places) (Kramer 1997). Among the surveyed villages, Pachpadra is a good illustration. Two sub-castes, the Purubiya and the Banda, each of them sub-divided into three gotras, co-exist, and a sample of 31 brides we could carefully investigate show that the brides come from 24 different villages, the majority located more than 60 km away. As a consequence, kinship connections within the Hindu community typically involve large marriage networks at a regional level but with no concentration of ties within a given village, and sparse and weak family relationships across villages (Rowe 1960). The absence of regular visits between families, which contrasts with the frequent encounters at family events within the Muslim community, testifies of distended relationships between wife-givers and wife-takers, with no reciprocal relationships and perhaps the desire in some cases for no interference from in-laws (Miller 1985).

Appendix 2. Methodological appendix

Appendix 2.1. Bridge width

To capture the width of the bridges, we implemented (in NetLogo) the following procedure: for each pair of potters i and j who are not directly related, (1) we select the direct (i.e., one-step) neighbors of i; (2) we select the direct (i.e., one-step) neighbors of j; (3) we select the neighbors of the neighbors of i (i.e., neighbors of i at distance 2); (4) we count how many neighbors of the neighbors of i are neighbors of j. This last operation indirectly makes it possible to retrieve the number of links between the neighborhoods of the two disconnected potters because, when two-step neighbors of i are also one-step neighbors of j, only one of the two cases is verified: either the one-step neighbor of i has a link to one of the one-step neighbor of j or the one-step neighbor of i is also a one-step neighbor of j (which is the case of the common neighbor appearing in Centola and Macy’s definition, 2007: 713).

Appendix 2.2. Data limitation—family networks

Complete family trees were directly reconstructed with all interviewees, and, when ties beyond the limits of the surveyed households were mentioned, they were systematically double-checked, sometimes through quick questions, with potters living in the village concerned by these declarations. Our ability to travel within the Jodhpur region over a period of 3 years thus considerably limited observation biases usually present in the mapping of kinship connections (for a deep analysis of these problems, see Hamberg et al. 2011; Roth et al. 2013; Hamberg and Gargiulo 2014). The only simplification we introduced in our analysis is that we did not distinguish between different types of family ties so that, for instance, a cousinship and a brother-in-law relationship are regarded as equally relevant for the circulation of information about the kiln. Our qualitative evidence was indeed insufficient to formulate more specific hypotheses in this respect.

Appendix 2.3. Bridge fraction

Unless we are mistaken, Centola (2015: 1318) did not make explicit how he computed the “fraction of neighborhoods in the network with at least one ‘wide bridge’ to another neighborhood.” More specifically, the problem is how to compute the denominator of this “fraction.” For us, the most coherent way to compute that “fraction” is by making the denominator the potential number of not directly connected pairs of nodes in the network, which corresponds to the total number of possible dyads (N * (N − 1)), minus the number of observed ties in the network. The computation of the total number of possible dyads does not include the usual division by 2 for symmetric ties for the following reason: although the kinship network is symmetric, the width of the bridge between a given pair of not directly connected nodes i and j is not necessarily the same for the two nodes. In particular, as long as the number of neighbors at distance 1 of i and j differs, the node with the smallest number of neighbors at distance 1 is structurally likely to have the narrowest bridge.

Appendix 2.4. Data limitation—advice networks

Compared to our family networks, it must be acknowledged that information about advice is more fragile. The advice networks we reconstructed are a rough approximation of the unobservable influence process to which potters were likely to be exposed. In this respect, let us first note that we did not have access to the complete ego-centered networks (of a randomly selected sample of potters) on several dimensions (for such an exceptional design, see Banerjee et al. 2013). Second, we did not have access to fine-grained time-stamped advice exchanges. For instance, when a potter is described as being invited to give a public demonstration attended by a certain number of potters, as in the excerpt above from Ramasani (see Appendix 1.1), we attribute to this potter the role of main “influencer,” though some of the potters attending the demonstration adopted several months (or years) later, thus suggesting that in fact interactions following the demonstration may (also) have played a role in the decision process. Because of these limitations—which indeed are common in network studies relying on historical or qualitative data (see, for instance, Böhm and Hillmann 2015: 168; Mitschele 2014; Hollstein 2011)—, we decided not to estimate complex multivariate statistical models for network configuration (Robins 2011) and/or evolution (Snijders 2011) or individual-level models with network covariates relating potters’ time of adoption to their positions in the advice network (van Duijn and Huisman 2011; see also Valente 1999: ch. 8).

Appendix 2.5. QAP correlations

Taking inspiration from analyses on multiplex networks (for two remarkable illustrations with respect to marriage and business ties, and advice and friendship connections, see, respectively, Padgett and Ansell 1993, and Lazega and Pattison 1999), we first quantify the proportion of pair of potters for which an advice and a family tie was observed at the same time. Among Indian-Muslim potters, 96% of advice ties correspond to a family tie; among Indian-Hindu potters, 60% of advice ties are also family ties; among Kenyan-Mukurino and other-religion potters (as before considered as a single network because mixed-religion diffusion ties are present), 55% of advice ties are also family ties. Then, quadratic assignment procedure (QAP) correlations were used to test for the robustness of these counts (for a description of this method, see Borgatti et al. 2013: 128–129; Robins 2015: 190–191). QAP correlations for Muslim, Hindu, and, on the other hand, Mukurino and other-religion potters are, respectively, 0.254 (p < 0.001), 0.247 (p < 0.001), and 0.08 (p = 0.07). If one considers the meaning of the p value within the quadratic assignment procedure—that is, the proportion of randomly permutated networks for which one finds the same correlation observed on the empirical networks—, then it seems reasonable to consider that, although the result is slightly less spectacular for Kenyan networks, it is highly unlike that the observed overlap in both social contexts between advice and kinship ties is due to chance or structural constraints only. (QAP correlations were computed through the function “qaptest” in R package “sna” and double-checked through a program we wrote in Java from scratch.)

Appendix 2.6. Simulated time and interaction rate (IR)

In our simulations, one iteration represents 1 day so that 6 months, the temporal unit over which our empirical diffusion rates were computed (see Figs. 1 and 2), amount to 180 iterations. The simulations last for 52 (Indian case) or 32 semesters, the temporal window for which we have empirical data, namely from 1987 (Indian case) or 1997 (Kenyan case) to the second half of 2013. To get the total number of interactions for a given model variant, it suffices to multiply the interaction rate by 180 (the number of days in one semester) by 52 (the total number of semesters over which the simulation unfolds). For simulations for the Indian case, when the interaction rate is equal to 1, 2, 3, or 4, this means that 1, 2, 3, or 4 adopter/potential adopter pairs interact at each iteration. When the interaction rate is equal to 0.25 or 0.5, this means that, at each iteration, there is only a probability of 0.25 or 0.5 that a pair of adopter/potential adopter gets in contact. For simulations for the Kenyan case, a lower number of adopter/potential adopter pairs to be sampled at each iteration is required because the Kenyan potter population is smaller, the diffusion process spans over a shorter while, and pottery only occupies a month a year. In particular, when the interaction rate is equal to 0.005, this means that, at each iteration, there is only a probability of 0.005 that a pair of adopter/potential adopters gets in contact, thus implying that, over a semester (180 iterations), only one interaction takes place. Following the same reasoning, readers can deduce that 0.02, 0.03, 0.05, 0.09, and 0.18 respectively correspond to 4, 6, 10, 16, and 32 interactions per semester.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Manzo, G., Gabbriellini, S., Roux, V. et al. Complex Contagions and the Diffusion of Innovations: Evidence from a Small-N Study. J Archaeol Method Theory 25, 1109–1154 (2018). https://doi.org/10.1007/s10816-018-9393-z

Download citation

Keywords

  • Ceramic techniques
  • Innovation diffusion
  • Weak ties
  • Strong ties
  • Kinship networks
  • Complex contagions
  • Agent-based simulations