Cultural Evolution Mechanisms

Disentangling how topology, selection, and reproduction interact in experimental cultural evolution.

Cultural evolution is rarely driven by a single force. Cultural change depends on how people are connected, which variants they choose to imitate, and how faithfully or creatively they reproduce what they observe. This project uses experimental social networks to characterize how these mechanisms interact.

Participants reproduce short melodies while embedded in different social network topologies. By manipulating network structure, selection, and reproduction, the experiment reveals how local interactions can generate population-level cultural structure, including shared melodic prototypes, diversity, and differences in aesthetic value.

Read the paper PDF

Paradigm for studying cultural evolution mechanisms in social networks
Schematic of the paradigm. (A) The cultural process is decomposed into three components: topology, selection, and reproduction. Participants listen to neighbors, choose a melody, and reproduce it. (B) The study compares a square lattice, a random regular graph, and a modular network.

The paradigm makes it possible to study cultural evolution as a system of interacting mechanisms rather than isolated effects. Network topology determines which cultural variants are available locally; selection determines which variants are preferentially chosen; reproduction determines how variants are transformed as they pass from person to person.

Emergent melodic prototypes in experimental social networks
Emergent melodic prototypes. (A) Melodic contours derived from jointly clustering all reproduced melodies, shown as deviations from the mean in semitones. (B) Mean z-scored pleasantness ratings for each melodic cluster. (C) Example evolution of prototypes over time from one experimental batch. Nodes are colored by melodic cluster, and edges are highlighted when neighboring nodes share the same cluster. Error bars show 95% confidence intervals.

The experiments show that topology can shape which melodic prototypes emerge and persist. Modular networks, lattice structures, and random graphs produce different local environments for imitation, which in turn influence convergence, diversity, and the spread of more or less pleasant variants.

Comparison of linear and networked cultural evolution
Comparison between linear and non-linear cultural evolution. (A) Cluster prevalence across iterations. (B) Clustered melodies in PCA space across the final iterations. (C) Average population pleasantness after a three-iteration burn-in period, with all conditions initialized randomly and identically. Error bars show 95% confidence intervals.

What we do

  • Decompose cultural evolution into interacting mechanisms of topology, selection, and reproduction.
  • Use networked melody transmission experiments to observe how local imitation scales into population-level cultural structure.
  • Compare linear transmission with networked transmission across lattice, random, and modular social structures.
  • Analyze emergent prototypes, pleasantness, entropy, and neighbor similarity over time.
No-selection condition in networked melody transmission
No-selection condition. (A) Melodies are sampled uniformly from the local environment for imitation. (B) Average population pleasantness after a three-iteration burn-in. (C) Evolution of neighbor similarity and entropy. Markers indicate experimental batches by condition, with more transparent colors indicating earlier iterations. (D) Average deviation from the mean for neighbor similarity and entropy after burn-in. Error bars show 95% confidence intervals.

Why it matters

Understanding cultural evolution requires knowing how mechanisms combine. Selection can favor appealing variants, reproduction can transform them, and topology can determine which variants meet and compete. This work provides an experimental framework for studying those interactions directly, helping explain when cultural systems converge, diversify, or settle into distinct local traditions.

(Marjieh et al., 2025)

Related Publications

2025

  1. Raja Marjieh, Manuel Anglada-Tort, Thomas L Griffiths, and 1 more author
    arXiv preprint arXiv:2502.12847, 2025