Question: How do descriptors of forest vegetation (species richness and composition) respond to edge influence and do these response patterns fit to the two-zone pattern theoretically predicted and empirically observed? Location: District of Aurignac, southwestern France. Methods: We surveyed all understory forest vegetation along 28 transects pertaining to seven contrasted boundary types. Boundary types were a combination of adjacent land cover, orientation and topography classes. Response patterns to edge influence of two vegetation descriptors, species richness and composition (summarized by an ordination), were analysed using the two-phase linear regression method. Data for each transect were divided into all possible pairs of adjacent groups of successive plots, and a linear regression was computed on both groups of plots. Regression parameters were interpreted by comparison with the two-zone theoretical pattern and related to boundary types. Results: We have shown for nearly half of the transects that vegetation responses deviate from the two-zone pattern. There was no significant relationship between regression coefficients for richness and boundary characteristics whereas regression coefficients for composition were partly related to adjacent land cover. Conclusions: Our results demonstrate a high variability of forest vegetation response patterns to edge that contrasts with the theoretical pattern usually reported in the literature. Forest history and edge dynamics are invoked to explain the observed vegetation patterns. We have shown the need for a review of factors affecting the edge influence for better management and conservation of forest plant species.