It is suggested that error analysis in elementary laboratories be carried out by computer simulation of repeated measurements. Results should be represented as histograms. This procedure is concrete and correct, and makes definite and graphic what is meant by "error analysis."