The Cauchy distribution p(x) = (a/π) (1/(x^2 + a^2)) can be realized. Erect a vertical screen along the x-axis. Pivot a laser pointer about the vertical axis through (0,a). Spin the laser and record the x-coordinate of the spot, or record the average intensity at x. These satisfy the Cauchy distribution. -- Gene On Saturday, March 10, 2018, 7:04:23 AM PST, Henry Baker <hbaker1@pipeline.com> wrote: We've all seen those balls-and-pegs mechanical models which produce good approximations to binomial distributions (and hence Gaussian distributions). Are there any simple mechanical models for Zipf's Law? Power law? Pareto distribution? Log-normal distributions? In particular, if one were trying to produce a power law distribution, it would be nice if it somehow involved the equivalent of a log-log plot (where the distribution becomes linear).