The evolutionary importance of hybridization in wild plants and animals has become increasingly widely recognized in the last decade. In practical terms, hybridization provides an exceptionally tough set of problems for conservation biologists. We illustrate this in a case study of two Carabidae species widely used to evaluate the impact of human activities on biodiversity. These two species live in a complex mosaic of sympatry/allopatry and are known to hybridize in controlled conditions. Hybridization has not been quantified in natural populations to date due to the lack of a simple set of phenotypic traits for identifying hybrids. We thus screened for hybrids in natural populations, by multilocus genotyping at nine microsatellite loci. A high level of genetic differentiation between these two taxa was observed, as shown by allelic frequency distributions. Two Bayesian assignment procedures without obligatory pure taxon references were used to infer different classes of hybrids (F(1), F(2) and backcrosses) and mixture proportions between the two species. A low level of hybridization (F(1) genotypes) was observed in natural populations, contrasting with results obtained in controlled conditions. A high level of introgression was, however, detected at three of 12 sites, as revealed by the detection of backcrossed genotypes. This interspecific gene flow was detected in a limited zone of the common geographical range of the two species and was not related to the pattern of sympatry/allopatry. We then considered the origin and repercussions of this introgression, based on intraspecific genetic diversity and geographical structure.