You don't take the death cap (Amanita phalloides) home for tea. This species, pictured here, is infamously poisonous, with many other mushrooms being toxic to a greater or lesser degree.
Thomas N. Sherratt, David M. Wilkinson and Roderick S. Bain have addressed two issues raised by the existence of poisonous mushrooms (Am. Nat. doi:10.1086/497399). The first question was what purposes possession of poisons might serve in mushrooms. One possibility is that toxins are simply a metabolic by-product. Another that has been suggested by several authors is that they act as a deterrent to predators, which might otherwise destroy the mushroom before its spores have matured and dispersed. Fungus-loving vertebrates could in particular be highly destructive.
An evolutionary principle is that if you as an organism go to the bother of being unpalatable, you might as well signal that fact. Does this apply in mushrooms? To investigate this second issue, Sherratt et al. turned to data compilation and neural-network analysis. They made use of modern evolutionary trees to judge the incidence of poisonousness in mushrooms, then analysed data sets, culled from field guides, to see whether poisonous species tend to have particular ecological correlates — whether, for instance, they are more colourful, more aggregated or have a more noticeable odour.
Overall odour (and not cap colour) came out as the best predictor of toxicity, a result that was supported by pairwise comparisons of related poisonous and edible forms. Given that many animals forage by night, and that nocturnal mammals tend to have relatively poor colour vision, the authors suspect that odour provides the more effective signal.
Sherratt et al. make plain that their study is correlative only, and that — for them and others — this is a work in progress. There is rich scope for further investigation of the hypothesis that poisonous mushrooms use odours as warning signals, and of the likely exceptions.