We tested the hypothesis that large herbivores manipulate their own food supply by modifying soil nutrient availability. This was investigated experimentally by determining the impact of faeces on grasses, mosses and soil biological properties in tundra ecosystems. For this, we increased the density of reindeer Rangifer tarandus platyrhynchus faeces and studied the response of a tundra system on Spitsbergen to this single faecal addition treatment for four subsequent years. From the third year onwards faecal addition had unambiguously enhanced the standing crop of grasses, as evidenced by an increase in both shoot density and mass per shoot. Although reindeer grazing across experimental plots was positively related to the abundance of grasses in anyone year, the increase in grass abundance in fouled plots failed to result in greater grazing pressure in those plots. Faecal addition enhanced soil microbial biomass C and N, particularly under wet conditions where faecal decay rates were greatest, whilst grasses appeared to benefit from faeces under dry conditions. Whilst growth of grasses and soil microbial biomass were stimulated by faecal addition, the depth of the extensive moss layer that is typical of tundra ecosystems was significantly reduced in fouled plots four years after faecal addition. The greatest reduction in moss depth occurred where fouling increased soil microbial biomass most, suggesting that enhanced decomposition of moss by a more abundant microbial community may have caused the reduced moss layer depth in fouled plots. Our field experiment demonstrates that by the production of faeces alone, vertebrate herbivores greatly impact on both above- and belowground components of tundra ecosystems and in doing so manipulate their own food supply. Our findings verify the assertion that grazing is of fundamental importance to tundra ecosystem productivity, and support the hypothesis that herbivory is instrumental in promoting grasses whilst suppressing mosses. The widely observed inverse relationship between grass and moss abundance in the field may therefore reflect the long history of plant-herbivore interactions in tundra ecosystems.