The effect of forest fragmentation was studied in hoverfly communities of 54 isolated forests (0.14–171 ha) in south west France. The positive relationship between species richness and wood patch area was investigated by testing the three hypotheses usually put forward to explain it: 1) the sampling effect hypothesis, 2) the patch heterogeneity hypothesis, 3) the hypothesis of equilibrium between distance from other patch (colonisation) and surface area of the patch (extinction). The syrphid species were divided into 3 ecological groups, based on larval biology as summarized in the "Syrph the Net" database: non forest species, facultative forest species and forest species. A total of 3317 adults belonging to 100 species, were captured in the 86 Malaise traps. Eight species were non forest (N=16), 65 facultative forest (N=2803) and 27 forest species (N=498). Comparison of the slopes of the species-area curves for species richness and species density per forest patch showed a strong sampling effect in the species-area relationship. Wood patch heterogeneity increased with wood patch area and positively influenced hoverflies richness. Less isolated wood patches presented high richness of forest species and low richness of non forest species. Only forest species richness seemed to respond to the equilibrium between surface area and isolation. Depending on which hypothesis explained best the species-area relationship, management recommendations to mitigate fragmentation effects were formulated at various spatial scales and for different stakeholders.