Skip to content
Back to search results

Multilayer Social Network Analysis

Introduced by
Kivela, M.; Boccaletti, S. et al.
Year
2014
Type
Structural network analysis framework
Family
Network science
Font size

What does it do?

Multilayer social network analysis extends classical single-layer network methods to settings where actors are connected through multiple, distinct types of ties — such as friendship, professional collaboration, and online interaction — simultaneously.

By modeling each type of relationship as a separate layer and explicitly representing connections across layers, it captures structural complexity that a single aggregated network would hide.

The intuition behind the method

In everyday life, people are embedded in overlapping networks: a colleague is also a friend, a Twitter contact, and a co-author. Collapsing all these ties into one graph loses information about which ties operate in which context. Multilayer analysis preserves these distinctions by stacking the relationship types as layers and linking each person's node across layers. This lets researchers ask questions like: which actors are central in every layer, or how does influence propagate from one type of tie to another?

History

The formal mathematical framework for multilayer networks was consolidated in a pair of influential 2014 papers by Kivela et al. and Boccaletti et al., which unified earlier work on multiplex networks, interdependent networks, and network-of-networks models. The approach grew from physics and complexity science and has since been widely adopted in sociology, organizational research, communication studies, and epidemiology, where multiple relationship types are the norm rather than the exception.

When to use it

Use multilayer social network analysis when actors are embedded in multiple, qualitatively distinct relationship types that you believe have independent structural effects, and when collapsing them into a single graph would obscure important variation. It is appropriate for datasets with clear layer identities — for example, online versus offline ties, friendship versus advice networks, or cross-platform communication. Prefer single-layer SNA when only one relationship type is measured, when the dataset is very small (fewer than about 30 nodes), or when the research question does not depend on distinguishing tie types. The method requires software that supports multilayer graph objects (e.g., MuxViz, R igraph with layer attributes, NetworkX-based tools) and researcher familiarity with multilayer centrality concepts.

Limitations and considerations

  • Requires complete or near-complete data across all layers; missing one layer introduces systematic bias.
  • Computational cost grows with the number of layers and nodes, making very large multilayer networks expensive to analyse.
  • Interpretation is more complex than single-layer SNA; communicating results to non-technical audiences requires care.
  • Standardised software and reporting conventions are still maturing compared to classical SNA.
  • Small networks (fewer than ~30 nodes) often produce unstable centrality and community estimates in the multilayer setting.

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Captures the full complexity of social life by preserving distinct relationship types rather than flattening them.
  • Enables comparison of structural positions across layers, revealing actors who are central in some contexts but peripheral in others.
  • Community detection across layers identifies robust social groupings that persist across multiple relationship types.
  • Interlayer coupling metrics quantify how strongly different relationship contexts co-determine social position.
  • Extendable to temporal and directed variants, making it applicable to a wide range of social science questions.

Applications

  • Organizational research examining advice, friendship, and reporting ties simultaneously among employees.
  • Online-offline social network studies linking platform-specific interactions to face-to-face relationships.
  • Epidemiological modelling of disease spreading across co-worker, household, and transport contact layers.
  • Political science research on coalition formation across legislative, media, and lobbying networks.
  • Science of science studies combining co-authorship, citation, and institutional affiliation layers.
  • Communication studies tracking information diffusion across multiple social media platforms simultaneously.

How it works

How it works

1

Each distinct type of social relationship is encoded as a separate layer: a graph whose nodes are the same set of actors but whose edges reflect only that relationship type. Layers may represent communication channels, temporal snapshots, or qualitatively different tie types such as advice-seeking versus emotional support.

2

Edges called interlayer couplings connect each actor's representation across layers, modeling the fact that the same person participates in multiple relationship contexts. Coupling can be categorical (same actor, different layer) or weighted to reflect the strength of cross-layer dependence.

3

Standard centrality measures — degree, betweenness, eigenvector — are generalised to the multilayer setting, producing layer-specific and aggregate scores. Measures such as multilayer PageRank or versatility allow comparison of node importance across the full system rather than within any single layer.

4

Community detection algorithms adapted for multilayer networks identify groups of actors who are densely connected within the same communities across multiple layers simultaneously. This reveals whether social groupings are consistent across relationship types or fragmented by context.

5

Results are interpreted both at the layer level and across the full multilayer system. Researchers examine which layers drive overall network structure, identify actors who bridge layers, and test whether patterns observable in one layer replicate in others.

Common pitfalls

  • Treating the choice of layers as inconsequential — layer definition is a substantive theoretical decision that shapes all downstream results.
  • Ignoring interlayer coupling strength and assuming independence between layers without empirical justification.
  • Using aggregate (collapsed) centrality as the sole result and omitting layer-specific analysis.
  • Failing to distinguish between multiplex networks (same nodes, different edge types) and multilevel networks (nodes at different hierarchical levels), which require different models.
  • Reporting community assignments without validating stability through permutation tests or robustness checks.

Frequently asked questions

What is the difference between a multilayer and a multiplex network?

A multiplex network is a special case of a multilayer network in which the same set of nodes appears in every layer and interlayer edges connect each node only to its own counterpart in other layers. Multilayer networks are more general and allow different node sets per layer and arbitrary interlayer connections.

What software can I use for multilayer social network analysis?

MuxViz provides a graphical interface for multilayer visualisation and centrality. The R packages igraph and multinet, as well as Python libraries such as pymnet, support multilayer network construction and analysis programmatically.

How do I decide which layers to include?

Layer selection should be driven by theory: include only relationship types that are conceptually distinct and relevant to the research question. Adding redundant or highly correlated layers rarely improves insight and increases analytical complexity.

Can multilayer SNA handle missing data in one layer?

Partial missingness can be accommodated by weighting layers or using imputation strategies, but systematic absence of an entire layer for a subset of actors introduces bias. Sensitivity analyses that exclude problematic actors or layers are strongly recommended.

Is multilayer centrality comparable across studies?

Not directly, because centrality values depend on the number and type of layers included. Always report the layer structure alongside centrality values, and compare relative rankings rather than raw scores across studies.

Related methods

Sources

  1. Kivela, M., Arenas, A., Barthelemy, M., Gleeson, J. P., Moreno, Y., & Porter, M. A. (2014). Multilayer networks. Journal of Complex Networks, 2(3), 203–271.

    Open DOI reference
    Copied.
  2. Boccaletti, S., Bianconi, G., Criado, R., del Genio, C. I., Gomez-Gardenes, J., Romance, M., Sendina-Nadal, I., Wang, Z., & Zanin, M. (2014). The structure and dynamics of multilayer networks. Physics Reports, 544(1), 1–122.

    Open DOI reference
    Copied.