The utility of the new polymerase chain reaction (PCR)-based multilocus DNA fingerprinting technique amplified fragment length polymorphism (AFLP) for paternity analysis in natural plant populations was assessed. In a natural population of 25 plants of Persoonia mollis (Proteaceae), three AFLP primer pairs generated 147 dominant loci. Of these, 125 (85%) were polymorphic, with a mean recessive allele frequency of 0.735. The theoretical expected percentage of offspring for which all males except the true father can be excluded (PET) was 99.9% for this population. The estimates of PET drop marginally to 99.6% and 97.6% for larger populations of 100 and 1000 individuals, respectively. A preliminary investigation confirmed the power of AFLP for paternity analysis by assigning paternity, or excluding all known potential sires, for 242 of 252 (96.0%) naturally pollinated seeds. Ambiguous paternity for the remaining 10 seeds was quickly resolved by utilizing two further AFLP primer pairs, ultimately generating over 200 polymorphic loci and resulting in the exclusion of all nonsires for all 252 (100%) seeds. This study highlights the utility of AFLP for paternity analysis because: (i) it generates sufficiently large numbers of highly reproducible polymorphic loci, that are (ii) quickly and accurately scored using an automated DNA sequencer and dedicated software, and (iii) unlike microsatellites, requires no sequence knowledge so it is more easily applied to new study species.