Figure 6.

Sample size required for finding an effect of a particular size (in milliseconds), derived by Monte Carlo simulation using two random samples of words from the database. For each combination of sample size (n = 10, 20, 40, 80, 160, 320, 640, 1280) and effect size (0, 5, 10, 20, 40 ms), we ran 1000 simulations. The y-axis indicates the proportion of simulations in which the null hypothesis (no-effect) was rejected. Sample sizes at which sufficient power (0.8) is reached are n = 33 for 40 ms, n = 126 for 20 ms, and n = 493 for 10 ms).