Random amplified polymorphic DNA (RAPD) fragment reproducibility was assayed in three animal species: red deer (Cervus elaphus), wild boar (Sus scrofa) and fruit fly (Drosophila melanogaster). Ten 10-mer primers (Operon) were tested in two replicate reactions per individual under different stringency conditions (annealing temperatures of 35 °C or 45 °C). Two estimates were generated from the data: autosimilarity, which tests the reproducibility of overall banding patterns, and band repeatability, which tests the reproducibility of specific bands. Autosimilarity (the similarity of individuals with themselves) was lower than 1 for all three species ranging between values of 0.66 for Drosophila at 45 °C and 0.88 for wild boar at 35 °C. Band repeatability was estimated as the proportion of individuals showing homologous bands in both replicates. The fraction of repeatable bands was 23% for deer, 36% for boar and 26% for fruit fly, all at an annealing temperature of 35 °C. Raising the annealing temperature did not improve repeatability. Phage lambda DNA was subjected to amplification and the pattern of bands compared with theoretical expectations based on nucleotide sequence. Observed fragments could not be related to expected ones, even if a 2bp mismatch is allowed. Therefore, the nature of genetic variation uncovered by the RAPD method is unclear. These data demonstrate that prudence should guide inferences about population structure and nucleotide divergence based on RAPD markers.