Pollinators such as bees provide a critical ecosystem service that can be impaired by information about predation. We provide the first evidence for olfactory eavesdropping and avoidance of heterospecific alarm signals, alarm pheromones, at food sources in bees. We predicted that foragers could eavesdrop upon heterospecific alarm pheromones, and would detect and avoid conspicuous individual pheromone compounds, defined by abundance and their ability to persist. We show that Apis cerana foragers avoid the distinctive alarm pheromones of A. dorsata and A. mellifera, species that share the same floral resources and predators. We next examined responses to individual alarm pheromone compounds. Apis cerana foragers avoided isopentyl acetate (IPA), which is found in all three species and is the most abundant and volatile of the tested compounds. Interestingly, A. cerana also avoided an odor component, gamma-octanoic lactone (GOL), which is >150-fold less volatile than IPA. Chemical analyses confirmed that GOL is only present in A. dorsata, not in A. cerana. Electroantennogram (EAG) recordings revealed that A. cerana antennae are 10-fold more sensitive to GOL than to other tested compounds. Thus, the eavesdropping strategy is shaped by signal conspicuousness (abundance and commonality) and signal persistence (volatility).

The nectar of the thunder god vine, Tripterygium hypoglaucum, contains a terpenoid, triptolide (TRP), that may be toxic to the sympatric Asian honey bee, Apis cerana, because honey produced from this nectar is toxic to bees. However, these bees will forage on, recruit for, and pollinate this plant during a seasonal dearth of preferred food sources. Olfactory learning plays a key role in forager constancy and pollination, and we therefore tested the effects of acute and chronic TRP feeding on forager olfactory learning, using proboscis extension reflex conditioning. At concentrations of 0.5-10 µg TRP ml-1, there were no learning effects of acute exposure. However, memory retention (1 h after the last learning trial) significantly decreased by 56% following acute consumption of 0.5 µg TRP ml-1 Chronic exposure did not alter learning or memory, except at high concentrations (5 and 10 µg TRP ml-1). TRP concentrations in nectar may therefore not significantly harm plant pollination. Surprisingly, TRP slightly increased bee survival, and thus other components in T. hypoglaucum honey may be toxic. Long-term exposure to TRP could have colony effects but these may be ameliorated by the bees' aversion to T. hypoglaucum nectar when other food sources are available and, perhaps, by detoxification mechanisms. The co-evolution of this plant and its reluctant visitor may therefore likely illustrate a classic compromise between the interests of both actors.

Social pollinators such as honey bees face attacks from predators not only at the nest, but also during foraging. Pollinating honey bees can therefore release alarm pheromones that deter conspecifics from visiting dangerous inflorescences. However, the effect of alarm pheromone and its chemical components upon bee avoidance of dangerous food sources remains unclear. We tested the responses of giant honey bee foragers, Apis dorsata, presented with alarm pheromone at a floral array. Foragers investigated the inflorescence with natural alarm pheromone, but 3.3-fold more foragers preferred to land on the 'safe' inflorescence without alarm pheromone. Using gas chromatography-mass spectrometry analysis, we identified eight chemical components in the alarm pheromone, of which three components (1-octanol, decanal and gamma-octanoic lactone) have not previously been reported in this species. We bioassayed six major compounds and found that a synthetic mixture of these compounds elicited behaviors statistically indistinguishable from responses to natural alarm pheromone. By testing each compound separately, we show that gamma-octanoic lactone, isopentyl acetate and (E)-2-decen-1-yl acetate are active compounds that elicit significant alarm responses. Gamma-octanoic lactone elicited the strongest response to a single compound and has not been previously reported in honey bee alarm pheromone. Isopentyl acetate is widely found in the alarm pheromones of sympatric Asian honey bee species, and thus alarmed A. dorsata foragers may produce information useful for conspecifics and heterospecifics, thereby broadening the effects of alarm information on plant pollination.

Pollinators provide a key ecosystem service that can be influenced by predation and predator avoidance. However, it was unclear whether pollinators can avoid predators by eavesdropping, intercepting predator signals. Using a natural species assemblage, we show that a bee can eavesdrop on and avoid the trail pheromone of a sympatric ant, while foraging on a native plant. The giant Asian honeybee, Apis dorsata, avoided Calliandra haematocephala inflorescences with live weaver ants, Oecophylla smaragdina. Although few foraging bees were attacked, ants killed the bee in almost a third of attacks. Ant presence alone significantly reduced bee floral visits. Bees showed nearly equal avoidance of live ants and trail pheromone extracts, demonstrating that olfactory eavesdropping alone can elicit full avoidance. We then used GC-MS to analyse compounds deposited by ants walking and laying trail pheromone. The most abundant compounds were all trail pheromone components. However, bees did not avoid the most abundant and conspicuous trail pheromone compound, heneicosane. Foragers may instead detect a mixture of different trail pheromone compounds. Our results contribute to a growing understanding of how public information about predators and competitors can shape food webs, and show that pollinators can tap into the private signals of predators and use this information to their advantage.

Fear can have strong ecosystem effects by giving predators a role disproportionate to their actual kill rates. In bees, fear is shown through foragers avoiding dangerous food sites, thereby reducing the fitness of pollinated plants. However, it remains unclear how fear affects pollinators in a complex natural scenario involving multiple predator species and different patch qualities. We studied hornets, Vespa velutina (smaller) and V. tropica (bigger) preying upon the Asian honey bee, Apis cerana in China. Hornets hunted bees on flowers and were attacked by bee colonies. Bees treated the bigger hornet species (which is 4 fold more massive) as more dangerous. It received 4.5 fold more attackers than the smaller hornet species. We tested bee responses to a three-feeder array with different hornet species and varying resource qualities. When all feeders offered 30% sucrose solution (w/w), colony foraging allocation, individual visits, and individual patch residence times were reduced according to the degree of danger. Predator presence reduced foraging visits by 55-79% and residence times by 17-33%. When feeders offered different reward levels (15%, 30%, or 45% sucrose), colony and individual foraging favored higher sugar concentrations. However, when balancing food quality against multiple threats (sweeter food corresponding to higher danger), colonies exhibited greater fear than individuals. Colonies decreased foraging at low and high danger patches. Individuals exhibited less fear and only decreased visits to the high danger patch. Contrasting individual with emergent colony-level effects of fear can thus illuminate how predators shape pollination by social bees.