Number of samples depends on the effect size. e.g. You only need n=1 to claim a dog can talk.