About the Method: Response Item Networks (ResIN)

Within Response Item Networks, nodes represent responses to survey questions, and a link between two response nodes means that both responses tend to be held simultaneously within the studied population (i.e. a positive correlation). The higher the correlation between two response nodes, the closer they are placed to one another. The main axis of the network then shows which responses cluster together and which do not, without presupposing clusters, social groups etc. For more information about the method, software tools or the researchers behind it, see the following links.

Website: www.resinmethod.net

Published

  • Carpentras, D., Lüders, A. & Quayle, M. (2022). Mapping the global opinion space to explain anti-vaccine attraction. Scientific Reports, 12, 6188. doi:10.1038/s41598-022-10069-3
  • Lüders, A., Carpentras, D., & Quayle, M. (2023). Attitude networks as intergroup realities: Using network-modelling to research attitude-identity relationships in polarized political contexts. British Journal of Social Psychology. doi:10.1111/bjso.12665
  • Carpentras, D., Lüders, A., & Quayle, M. (2024). Response Item Network (ResIN): A network-based approach to explore attitude systems. Humanities and Social Sciences Communications, 11(1), 1-14. doi:10.1057/s41599-024-03037-x
  • Chen, Y., Speer, A., De Bruin, B., Carpentras, D., & Warncke, P. (2025). A “broken egg” of US political beliefs: Using response-item networks (ResIN) to measure ideological polarization. Network Science, 13, e20. doi:10.1017/nws.2025.10016
  • Warncke, P., Chen, Y., Speer, A., de Bruin, B., Lüders, A., & Carpentras, D. (2026). ResIN: A new method to analyze socio-political attitude systems. In Computational Social Science of Social Cohesion and Polarization (pp. 105-129). Springer. doi:10.1007/978-3-032-01373-6_5

Preprints and reports

  • Warncke, P., Carpentras, D., Lüders, A., & Quayle, M. (2026). Introducing ResIN: An applied tutorial to the Response-Item-Networks package for R. doi:10.7910/DVN/WQLSIP
  • Carpentras, D., Lüders, A., Warncke, P., & Morselli, D. (2026). Belief Network Theory: A conceptual framework for studying complex attitude-identity systems. doi:10.13140/RG.2.2.35745.88169
  • Carpentras, D. (2021). Bridging networks with linear spaces to study attitudes and political polarization [ResIN on Twitter data]. Link
  • Carpentras, D. (2019). A method for estimating distance in opinion dynamics models using surveys data [“proto-ResIN”]. doi:10.13140/RG.2.2.29644.77449

Comparing Belief Networks across Surveys

Made by: Bart de Bruin

Build time series by picking a survey, one or more countries (for the ESS and EVS) and the issue items, then press Update figures. The five ResIN metrics are plotted for every year in which all chosen items were asked. Click a year in any chart to draw the networks of all series for that year below.

Extreme Distance iExtreme Distance
Average distance between the two extreme answers of each issue (e.g. 'strongly agree' and 'strongly disagree'), relative to the widest distance in the network. 0 = the extremes lie together; 1 = they lie as far apart as the network is wide.

Separation iSeparation
Separation between the two clusters of extreme responses. For each issue, the two extremes are sorted into two opposing camps by maximizing modularity. The final score compares how strongly the extremes are correlated within their own camp with how strongly they are correlated with the other camp, relative to chance. 0 = no camps, 0.5 = two fully separated camps.Chen et al. (2025), Network Science. doi:10.1017/nws.2025.10016

Link Density iLink Density
Average strength of cross-issue correlations. Higher values = stronger overall constraint between political attitudes.Chen et al. (2025), Network Science. doi:10.1017/nws.2025.10016

Linearization iLinearization
How 'flat' or one-dimensional the belief network is. High values indicate attitudes collapse into a single left-right axis.Chen et al. (2025), Network Science. doi:10.1017/nws.2025.10016

Neutral Displacement iNeutral Displacement
How far neutral (middle) response nodes deviate from the center (x = 0). Higher = neutral opinions pulled toward one side.

Networks

Metrics: median over 100 Fruchterman-Reingold layouts per network (the single-survey apps use 500, so their values can differ slightly). Item names, wording, recoding and data sources are documented in each survey's own app (menu top left).