Taste rejection of potential food items is commonly observed in the natural world: the consumer will sample a prey item orally before rejecting it. It is generally assumed that such prey are chemically defended with toxins, and that these toxins are signalled to potential consumers by an aversive taste (e.g. Bowers 1980; Jarvi, Sillen-Tullberg & Wiklund 1981; Wiklund & Jarvi 1982; Sillen-Tullberg 1985). This linkage between defence and a chemical advertisement is considered advantageous to both parties. The prey benefits from advertising its toxicity because the consumer is warned early in the predation sequence before it has irretrievably injured the prey. The predator benefits if the aversive taste reliably informs it of the danger of ingesting something dangerous and/or costly. But, how reliable is aversive taste as a signal that a potential meal is truly toxic? Evolutionary ecologists often implicitly believe in this reliability, and 'bitter' or 'aversive' are often taken as synonyms for 'toxically defended' (e.g. Fisher 1930; Cott 1940; Edmunds 1974; Alatalo & Mappes 1996; Tullberg & Hunter 1996; Tulberg, Leimar & Gamberale-Stille 2000; Ruxton, Speed & Sherratt 2004). However, there is no strong evidence in support of this belief, and we argue that distasteful nontoxic prey items may be widespread, and that further consideration and experimental testing of the link between toxic defence and aversive taste is required. Some compounds would be dangerous if consumed. One of the functions of an animal's taste sensory system is to detect such compounds. However, as there are very many potentially toxic compounds, it is not practical for an animal to have dedicated taste receptors specific to each potential threat. Such a system would be physically impossible to accommodate and would be unfeasibly expensive to maintain. Hence, some degree of compromise is required; this compromise may be open to exploitation by prey. If some toxic compounds are highly unlikely to be encountered, then there is no need to carry the ability to detect such compounds. Further, as developing and maintaining taste receptors, and the means to process information from them, are likely to incur some costs, we would expect redundant receptors to be selected out. However, this solution would leave the consumer vulnerable to the emergence of a novel toxin in their prey, with this toxin being consumed without warning, until such times as an appropriate receptor or behavioural avoidance evolves. Another solution for consumers is to have receptors that are not specific to a single chemical compound but rather are triggered by a range of compounds. This solution requires a trade-off between specificity and generality. Generality is made more attractive by the fact that the consumer generally does not gain a fitness benefit from identifying a specific toxic compound, it may be sufficient merely to identify potential prey as dangerous to eat or not. If a particular receptor is too general in the compounds that trigger it, then it will be triggered by nontoxins as well as toxins and the consumer will mistakenly taste-reject benign and nutritious meals. However, if the receptors are too specific then either an unfeasible number of different receptors would be required, or the consumer would have to run the risk of being unable to detect some toxins. For a specific consumer, we would expect the number and specificity of taste receptors to be shaped by evolution to trade-off the risk of failing to taste-detect toxins, and the risk of mistakenly rejecting undefended food items. The best solutions to this trade-off will differ between species, but the trade-off will always exist, and so no perfect cheat-proof taste detection of toxins will have evolved. The potential is there for consumers to be fooled into finding harmless compounds aversive. Do we expect selection pressure on potential prey to adopt unpalatable but otherwise benign compounds instead of truly toxic ones? Toxins are generally expensive either to synthesize or sequester (see review in chapter 5 in Ruxton et al. 2004), but the same could be true of aversive nontoxins. Further, it is not clear that one category of chemicals would be more or less expensive to manufacture or collect than the other. However, toxins are expensive in another way: the physiology of the bearer of the toxins has to be modified so as to avoid the animal itself being poisoned by the toxins. These modifications will likely have fitness costs (see Tollrian & Harvell 1999). These costs could be saved by instead storing a nontoxic compound that is aversive. There is evidence that compounds that fool taste receptors exist. Some fruiting plants produce hypersweet peptides that are cheaper to produce than the Correspondence: Graeme D. Ruxton, Division of Environmental & Evolutionary Biology, Institute of Biomedical & Life Sciences, University of Glasgow, Glasgow G12 8QQ, UK. E-mail: G.Ruxtongbio.gla.ac.uk ? 2006 The Authors. Journal compilation ? 2006 British Ecological Society