Fig 1.
Frequency of a focal allele over 100 generations. Each line shows evolution in a separate population of size N = 10 (A), N = 100 (B), N = 1000 (C) or N = 10000 (D). All simulations start at the same initial allele frequency of 0.5.
Fig 2.
Variance in cultural influence (; 2A), effective population size (Ne; 2B), mean Simpson diversity (2C) and mean number of unique cultural variants (on log scale; 2D) for different numbers of cultural role models R (with census population size N = 1000). Analytical results were confirmed by stochastic simulations. Grey diamonds in plot 2B show means and 90% prediction intervals for 1000 independent simulations (with μ = 10−4).
Fig 3.
Frequency-dependent transmission.
Effective population size (including 90% PIs) for anti-conformist (θ = 0.5; left), unbiased (θ = 1; center) and conformist transmission (θ = 1.5; right) and different innovation rates μ. Plots show trajectories for 100 generations after switch in transmission mode (1000 independent simulations; N = 1000).
Fig 4.
Variance (green squares) and inbreeding (yellow circles) effective population sizes (including 90% PIs; top), Simpson diversity (center) and mean numbers of unique variants (on log scale; bottom) for different migration rates m on the left and cultural exchange rates e on the right. We need to differentiate between effective size formulations because population sizes might differ between parental and offspring generations. Results come from 1000 independent stochastic simulations with census population size N = 1000.
Fig 5.
Exemplary networks (A-C), effective population sizes (with 90% PIs; D-F), mean numbers of unique cultural variants (on log scale; G-I) and Simpson diversity indices (J-L) for random (Erdős-Rényi), scale-free (Barabási-Albert) and small-world (Watts-Strogatz) networks. Parameter p gives the probability any two nodes are connected in random networks, π is the power of preferential attachment creating scale-free networks and K represents the number of initial neighbors on each side in small-world networks (with pr = 0.01). Note that because of structural differences between network types, the ranges of parameter values on the x-axes are not directly comparable. All graphs are created using the igraph R package [43]. Results come from 1000 independent simulations with census population size N = 1000. On the top, only 100 nodes are drawn for ease of illustration with p = 0.1, π = 1 and K = 4.