Fig. 2
From: A neural network model for the evolution of reconstructive social learning
Evolved time spent on unguided or self-guided individual learning in relation to the frequency and magnitude of environmental change. For three frequencies (f) and magnitudes (m) of environmental change, the panels show the number of learning episodes that evolved within 20 K generations in the cases of unguided and self-guided individual learning. A small number of learning episodes indicates that learning did not evolve. Each panel shows the outcome of 60 replicate simulations per learning implementation (unguided vs. self-guided individual learning). Each point represents the population mean averaged over the last 2 K generations of one simulation. The colour of each point represents the average performance in the population (‘performance’ = the quality of food items chosen, relative to the maximal possible quality). Artificial noise is added to the x-axis to improve readability. In all simulations, the width of the environmental profile was σ = 0.25 and social learning was not allowed to evolve.
