Effect of Network Structure on the Spread of Linguistic Behavior

Theoretical and computational models predict that social-network topology shapes the diffusion of linguistic forms, yet direct experimental evidence remains limited. Across three laboratory experiments, we tested whether network structure or seed placement influences the spread of target lexical items, while holding task demands constant. Experiment 1 compared clustered (high-transitivity) and random (low-transitivity) 12-node networks (N = 120). Experiment 2 extended this design to larger 24-node networks (N = 240). Experiment 3 provided an exploratory follow-up on seed placement within clustered networks, seeding either a dense local cluster or evenly spaced nodes (N = 120). In each study, participants engaged in dyadic problem-solving and image-description tasks while confederates covertly introduced target lexical items (e.g., antennae) to initiate diffusion. Lexical adoption was measured as the proportion of naming choices containing the target words, and lexical diversity was indexed using type-token ratio (TTR) across interaction rounds. Across all conditions, TTR declined linearly over time, indicating convergence on previously produced lexical items. Mixed effects logistic regression revealed no reliable differences in adoption between clustered and random network topologies or between seed-placement conditions. Together, these results suggest that, under the present conditions, lexical diffusion in small networks engaged in dyadic problem-solving is largely insensitive to global network structure or seed configuration. The findings identify boundary conditions on network effects in lexical alignment and highlight the need for richer interactional contexts and larger-scale networks in future experimental work.