Hybrids have long been recognized as a potential pathway for gene flow between species that can have important consequences for evolution and conservation biology. However, few studies have demonstrated that genes from one species can introgress or invade another species over a broad geographic area. Using 35 genetically mapped restriction fragment length polymorphism (RFLP) markers of two species of cottonwoods (Populus fremontii x P. angustifolia) and their hybrids (n = 550 trees), we showed that the majority of the genome is prohibited from introgressing from one species into the other. However, this barrier was not absolute; Fremont cpDNA and mtDNA were found throughout the geographic range of narrowleaf cottonwood, and 20% of the nuclear markers of Fremont cottonwood introgressed varying distances (some over 100 km) into the recipient species' range. Rates of nuclear introgression were variable, but two nuclear markers introgressed as fast as the haploid, cytoplasmically inherited chloroplast and mitochondrial markers. Our genome-wide analysis provides evidence for positive, negative, and neutral effects of introgression. For example, we predict that DNA fragments that introgress through several generations of backcrossing will be small, because small fragments are less likely to contain deleterious genes. These results argue that recombination will be important, that introgression can be very selective, and that evolutionary forces within the hybrid population to effectively "filter" gene flow between species. A strong filter may make introgression adaptive, prevent genetic assimilation, lead to relaxed isolating mechanisms, and contribute to the stability of hybrid zones. Thus, rather than hybridization being a negative factor as is commonly argued, natural hybridization between native species may provide important genetic variation that impacts both ecological and evolutionary processes. Finally, we propose two hypotheses that contrast the likelihood of contemporary versus ancient introgression in this system.