Many publications leave it to readers to guess whether reported data are based on independent biological replicates (e.g. number of animals) or technical replicates (e.g. one of several single measurements in an animal) as basis for reported means and statistical analysis. In a recent article, David Eisner explains that there are two sources of variability in a study in which 5 cells (or tissues) have been tested from 3 animals per group: the variation between the animals and that among cells isolated from an animal. Basing statistical analysis on the 3×5 cells/samples is wrong because it is a fundamental assumption of t-tests and ANOVA that samples are independent from one another; they are not if multiple samples come from the same animal. Violation of this independence rule can lead to artificially low p-values at the price of an inflated type I error (false positive). Modeling data are provided to demonstrate how number of animals vs. number of cells affect the percentage of experiments with p < 0.05. David Eisner emphasizes that there is nothing wrong with basing statistics on the number of cells in experiments when one is not comparing between animals. While technical replicates should not be used as the basis of statistical analysis, they are nonetheless highly recommended because their means/medians lead to more precise estimates of the value of an animal. This reduces variability and thereby increases the statistical power with a given number of animals.