Fig. 2

Step size distribution was closer to a Lévy-like random walk than a Brownian or patch-wise distribution. (A) Three hypothetical types of terrains with a patch-wise (left), a scale-free (middle), and a uniform (right) distribution of potential food sources. Three yellow dots show locations of hidden food pieces, which, in theory, could be identical across many distributions of potential food sources, among which three are shown here. This illustration suggests that when food is hidden, a hypothetical foraging path may be biased toward patch-wise (blue line) or Levy-like (red line) exploration merely by the distributions of potential food sources, independent of the actual location of the hidden food. In contrast, the uniform distribution minimizes this bias because it equally allows patch-wise (blue), Lévy-like (red), or Brownian (green) foraging paths. (B) Hypothetical distributions of step sizes for a patch-wise, a Lévy-like, or a Brownian walk. (C) The frequency of occurrences of step size, pooled across sessions of each monkey, binned into 20 cm bins. All distributions were unimodal (p = 0.09 (Vin;n = 324), 0.067 (Luk; n = 147), 0.13 (Hum; n = 148), and 0.07 (Nat; n = 202); Hartigan’s dip test). (D) same as B but shown on double log scales and normalized as a probability density function.