Ecosystems consist of interacting organisms. But interactions are assumed not to change. This is not necessarily so. In a simple saprophage community we tested the hypothesis that interactions are unvarying. The mould Aspergillus can produce toxins that kill Drosophila larvae. Drosophila inhibit mould. The mould can over grow yeast and yeasts are both damaged and facilitated by Drosophila. Paradoxically, larvae accumulated around both yeast and mould patches even though killed by mould. However larvae suvived in the 3 species combination. This effect was retained even when only yeast volatiles were present in the mould and larvae combination. The effect disappeared if the yeast grew on sugar-poor substrates. Metabolomics showed that mould in contact with volatiles from yeast growing on sugar-rich substrate was not producing many toxins. The expression of genes involved in producing toxins was suppressed by such volatiles.

Third party interactions therefore influence ecosystems by changing and even inverting interaction strengths. Interaction research must therefore accommodate these effects. They must also be included in predictions and models of ecosystem function.