Aims Coexistence of woody and herbaceous species in mixed communities has been an ecological conundrum for several decades. The general conclusion is that no single mechanism can explain coexistence. Researchers have proposed a unifying theory predicting that, despite states of equilibrium or domains of attraction, the mixed condition results from buffering mechanisms acting at the species level as the community approaches the limits of its existence (i.e., dominated by a single life-form). Here we experimentally tested, for the first time, these buffering mechanisms in grass and shrub species at the scale of a patch mosaic in a temperate shrub-grass steppe.Location Shrub-grass Patagonian steppes, Argentina (45 degrees 24 ' S, 70 degrees 17 ' W).Methods In six grazed and ungrazed plots large enough to represent accurately the vegetation mosaic of patches (121 m2), we created experimental communities exclusively dominated by grasses or shrubs. In addition, we established two control treatments: one with intact vegetation and another completely cleared. We constructed an integral projection model for populations of grasses and shrubs and evaluated the action of buffering mechanisms using the population growth rate over the transient period (lambda g).Results We highlight three main results. First, we found that lambda g of the reduced life-form increased when it was experimentally reduced. Second, there were species-specific differences that modulate the population response to dominance and disturbance (i.e., physiognomy and grazing, respectively). Third, fecundity was the vital rate driving the population response of both grasses and shrubs. Moderate grazing did not alter these results.Conclusions Buffering mechanisms play a critical role in the persistence of the co-dominated state of the shrub-grass community. We argue that buffering mechanisms act through processes promoting the fecundity of species reduced at the extremes by (i) negative intradominant life-form interactions (competition), (ii) niche partitioning that reduces inter-life-form niche overlap, and (iii) positive inter-life-form interactions (facilitation).