PCR-based multi-locus DNA fingerprints represent one of the most informative and cost-effective measures of genetic diversity and are useful population-level biomarkers of toxicologic and other anthropogenic impacts. However, concerns about reproducibility of DNA fingerprints have limited their wider use in environmental biology. We assessed polymorphism and reproducibility of two common fingerprinting techniques, RAPD (randomly amplified polymorphic DNA) and AFLP (amplified fragment length polymorphism), in pedigreed populations of rainbow trout (Oncorhynchus mykiss) to derive general rules for selective removal of problematic fingerprint bands. We found that by excluding bands that comprised less than 1% of total intensity, and by excluding the largest and smallest 10% of the bands, we could achieve nearly 100% reproducibility of AFLP fingerprints. Similar application of band exclusion criteria to RAPD fingerprints did not significantly enhance their reproducibility, and at least 15% of RAPD bands were not fully repeatable, heritable, or transmittable. The RAPD technique produced more polymorphic fingerprints than AFLP; however, considering that a substantial proportion of RAPD markers did not demonstrate Mendelian inheritance patterns, the AFLP methodology is to be preferred for future research.