The biodiversity impacts of non-native species should not be extrapolated from biased single-species studies

The presence, diversity and abundance of non-native plant species in natural vegetation are common condition indicators used to determine conservation status, with consequences for management strategies and investment. The rationale behind non-native species metrics as condition indicators is the assumption that non-natives have negative consequences on native biodiversity and habitat condition. The case against non-native species is not so clear-cut, with some studies reporting neutral or even facilitative interactions, often depending on spatial scale. Observational and experimental evaluations of the impact of particular non-native species on biodiversity provide a vital evidence-base for general conservation management strategies. Unintentionally though, many studies that quantify the impacts of non-native species have resulted in a publication bias in which species with known impacts are selected for investigation far more often than benign species. Here we argue that meta-analyses of the impacts of individual non-native species on natives, no matter how meticulous or objective, should not be generalized beyond the set of 'training' species. The likelihood of such extrapolation is increased when meta-analyses are reported with little qualification as to the skewed sampling towards problematic species, and because alternative findings such as non-native assemblages having positive interactions with native biodiversity, are under-reported. To illustrate, we discuss two meta-analyses that make general conclusions from impact studies skewed towards 'transformers', the most extreme invaders. We warn that if generic non-native species management strategies were to be based on these conclusions, they could not only fail to meet objectives but in some instances harm native biodiversity.