Mutualisms are interspecific interactions in which both players benefit. Explaining their maintenance is problematic, because cheaters should outcompete cooperative conspecifics, leading to mutualism instability. Monoecious figs (Ficus) are pollinated by host-specific wasps (Agaonidae), whose larvae gall ovules in their “fruits” (syconia). Female pollinating wasps oviposit directly into Ficus ovules from inside the receptive syconium. Across Ficus species, there is a widely documented segregation of pollinator galls in inner ovules and seeds in outer ovules. This pattern suggests that wasps avoid, or are prevented from ovipositing into, outer ovules, and this results in mutualism stability. However, the mechanisms preventing wasps from exploiting outer ovules remain unknown. We report that in Ficus rubiginosa, offspring in outer ovules are vulnerable to attack by parasitic wasps that oviposit from outside the syconium. Parasitism risk decreases towards the centre of the syconium, where inner ovules provide enemy-free space for pollinator offspring. We suggest that the resulting gradient in offspring viability is likely to contribute to selection on pollinators to avoid outer ovules, and by forcing wasps to focus on a subset of ovules, reduces their galling rates. This previously unidentified mechanism may therefore contribute to mutualism persistence independent of additional factors that invoke plant defences against pollinator oviposition, or physiological constraints on pollinators that prevent oviposition in all available ovules.

In some mutualisms, cooperation in symbionts is promoted by hosts sanctioning "cheats," who obtain benefits but fail to reciprocate. In fig-wasp mutualisms, agaonid wasps pollinate the trees (Ficus spp.), but are also exploitative by using some flowers as larval food. Ficus can sanction cheats that fail to pollinate by aborting some un-pollinated figs. However, in those un-pollinated figs retained by trees, cheats successfully reproduce. When this occurs, wasp broods are reduced, suggesting sanctions increase offspring mortality within un-pollinated figs. We investigated sanction mechanisms of abortion and larval mortality against wasp cheats in the monoecious Ficus racemosa by introducing into figs 1, 3, 5, 7, or 9 female wasps (foundresses) that were either all pollen-laden (P+) or all pollen-free (P-). The abortion rates of P- figs were highest (-60%) when single foundresses were present. Abortion declined with increased foundresses and ceased with seven or more wasps present, irrespective of pollination. In un-aborted figs, wasp fitness (mean offspring per foundress) declined as foundress number increased, especially in P- figs. Reduced broods in P- figs resulted from increased larval mortality of female offspring as foundress number increased, resulting in more male-biased sex ratios. Overall sanctions estimated from both abortion rates and reduced offspring production strengthened as the number of cheats increased. In a second experiment, we decoupled pollination from wasp oviposition by introducing one pollen-free foundress, followed 24 h later by seven pollen-laden ovipositor-excised wasps. Compared with P+ and P- single-foundress figs, delayed pollination resulted in intermediate larval mortality and wasp fitness, which concurred with patterns of female offspring production. We conclude that fig abortion reflects both pollinator numbers and pollen presence. Sanctions within P- figs initiate soon after oviposition and discriminate against female offspring, thus reducing the benefits to cheats from adaptively biasing their offspring sex ratios. We suggest that costs to cheats via these discriminative sanctions are likely to promote stability in this mutualism.

Fig trees are pollinated by fig wasps, which also oviposit in female flowers. The wasp larvae gall and eat developing seeds. Although fig trees benefit from allowing wasps to oviposit, because the wasp offspring disperse pollen, figs must prevent wasps from ovipositing in all flowers, or seed production would cease, and the mutualism would go extinct. In Ficus racemosa, we find that syconia (‘figs’) that have few foundresses (ovipositing wasps) are underexploited in the summer (few seeds, few galls, many empty ovules) and are overexploited in the winter (few seeds, many galls, few empty ovules). Conversely, syconia with many foundresses produce intermediate numbers of galls and seeds, regardless of season. We use experiments to explain these patterns, and thus, to explain how this mutualism is maintained. In the hot summer, wasps suffer short lifespans and therefore fail to oviposit in many flowers. In contrast, cooler temperatures in the winter permit longer wasp lifespans, which in turn allows most flowers to be exploited by the wasps. However, even in winter, only in syconia that happen to have few foundresses are most flowers turned into galls. In syconia with higher numbers of foundresses, interference competition reduces foundress lifespans, which reduces the proportion of flowers that are galled. We further show that syconia encourage the entry of multiple foundresses by delaying ostiole closure. Taken together, these factors allow fig trees to reduce galling in the wasp-benign winter and boost galling (and pollination) in the wasp-stressing summer. Interference competition has been shown to reduce virulence in pathogenic bacteria. Our results show that interference also maintains cooperation in a classic, cooperative symbiosis, thus linking theories of virulence and mutualism. More generally, our results reveal how frequency-dependent population regulation can occur in the fig-wasp mutualism, and how a host species can ‘set the rules of the game’ to ensure mutualistic behavior in its symbionts.