For example, when native pollinator populations have been reduced due to habitat fragmentation or other stressors, honey bees can “rescue” plants from reproductive failure , and, after honey bees have become naturalized, removing them may disrupt pollination of plants they would otherwise visit . However, regardless of whether honey bees are native or naturalized, dramatic increases of any species could disrupt species interactions and ecological processes , particularly when floral resources are limited. For example, in France, where honey bees are native, highly abundant managed honey bees can over-exploit limited floral resources, reducing pollen and nectar collection by wild bees . Indeed, although we studied only one plant species in a specific context, there are likely many systems for which introducing honey bees or other highly abundant generalist pollinators may indirectly reduce pollination by competitively displacing other pollinators. Several recent meta-analyses have revealed that honey bees are less effective than other bees . Furthermore, honey bees have been implicated in the extirpation of native bee species and frequently compete with other pollinators for limited pollen and nectar resources . Hive density is negatively correlated with wild bee abundance and diversity in many ecosystems and honey bees are replacing wild bees as floral visitors in some areas . Thus, indirect negative effects of honey bee introductions may be common where wild pollinator communities already effectively pollinate native plants. Conclusions – Our findings bear on ongoing discussion about permitting of honey bee hives on public lands.

Historically, the placement of managed hives in U.S. National Forests and Parks has been restricted and tightly regulated. However, vertical farm tower beekeepers have successfully lobbied to have honey bees considered a “non-consumptive” use of U.S. National Forest land . If adopted widely, such changes will likely lead to a massive increase in the number of managed honey bees in natural areas. Although honey bees are important pollinators in other systems, we show that indirect negative effects of competition can lead to overall negative effects of honey bee introductions on pollination. As such, introducing hives to sensitive ecosystems should be approached with extreme caution. More fundamentally, we show that introduced pollinators can disrupt plant-pollinator mutualisms and impair ecosystem functioning. These mutualists, although infrequently studied in the invasive species literature, broadly meet the definition of an “invasive” species despite their economic benefits to human society. Untangling direct and indirect effects allowed us to mechanistically understand the functional consequences of honey bee introductions. We recommend that future studies carefully consider indirect impacts of introduced species as biodiversity continues to decline and ecological communities become increasingly homogenous.Over 70% of plants depend to some degree on animal pollinators to successfully reproduce . Among the diversity of pollinators, taxa vary in their contributions to pollination in multiple intricate dimensions, some quantitative , others qualitative . At its core, the functional contributions of different pollinator taxa can be measured by the quantity and quality of visits to plant reproductive success . From a quantitative perspective, although biodiverse pollinator assemblages increase pollination , a few dominant species often provide the majority of floral visits . For example, the numerical dominance of honeybees as floral visitors has been hypothesized to drive their functional importance as pollinators . However, high visit frequencies can impair pollination in some contexts and we know little about whether strongly dominant visitors, such as honeybees, effectively pollinate the plants they visit.

Pollination effectiveness is defined as the per-visit contribution of floral visitors to pollination . A long history of studies within the botanical and evolutionary ecology literature documents variation in single visit effectiveness among plant visitors . To some extent, variation in pollination effectiveness reflects the wide range of methods used to measure it , such as single visit pollen deposition , the number of developed pollen tubes within styles , and/or fruit or seed set . Regardless, evidence for variation in SVE comes from numerous individual studies and this literature has yet to be synthesized in a way that would address whether and why particular taxa are more effective than others and whether dominant visitors are more effective pollinators of the plants they visit. Meta-analysis is a particularly valuable way to investigate such questions. An extensive literature on pollinator importance – the product of per-visit effectiveness and relative visitation rates of different pollinators – has concluded that pollinators that visit more frequently are generally more important . This conclusion suggests that numerical dominance outweighs among-species variation in SVE, but it is also possible that pollination effectiveness and visitation frequencies are correlated. First, frequent pollinators could be inherently more effective because of deep phylogenetic signals. For example, Ballantyne et al. found a positive correlation between a pollinator’s visit frequency and pollination effectiveness when comparing 23 plant species, likely because bees were both highly effective and highly frequent visitors compared to other floral visitors. Second, positive correlations between pollination effectiveness and visit frequency could occur if pollinators that visit frequently do so to the exclusion of other plant species. Such temporary fidelity or long-term fidelity would operate to minimize heterospecific pollen transfer, resulting in more effective pollination . On the other hand, high visitation rates may be the result of many quick and ineffective visits and have a negative or non-significant effect on reproductive success in many contexts . Despite their high visitation frequencies, the effectiveness of honeybees relative to other pollinators remains unclear. Bees are often the most effective pollinators of flowers and Apis mellifera is the most common flower-visiting bee species. However, there are several reasons to suspect that honeybees might be less effective than other bees. First, outside of their native range, honeybees lack the evolutionary history with endemic plants that could have selected for increased pollinator effectiveness . Furthermore, honeybees are floral generalists that visit a high proportion of available plants in ecosystems across the globe , and thus may not be particularly effective at pollinating specific flowering species. Second, honeybees sometimes ‘rob’ plants and efficiently extract and groom pollen from plants without depositing the pollen they extract or collect nectar without contacting reproductive structures . On the other hand, honeybees can be highly effective pollinators, even for plants with which they have no shared evolutionary history , suggesting that honeybees are highly adaptable and capable pollinators. Understanding pollinator effectiveness has important practical implications for safeguarding the production of pollinator-dependent crops. Highly effective non-honeybee pollinators are important for ensuring crop pollination in the face of global change and functionally diverse pollinator communities can increase crop pollination . Furthermore, pollination may differ in cultivated settings because interspecific plant competition, the spatial arrangement of flowers, and the pollinator taxa that provide pollination may vary between agricultural and natural landscapes .

We used a meta-analysis of the pollination effectiveness literature to address three key questions. First, how does the SVE of honeybees compare to that of other floral visitors? We hypothesized that honeybees would exhibit lower SVE relative to other pollinators because honeybees are broad generalists and might efficiently extract nectar and pollen without effectively pollinating plants. Second, to what extent do plant and pollinator attributes predict the comparative SVE of honeybees? Specifically, we evaluated whether pollinator taxonomic groups , crop status , and if plant species exist within the native range of honeybees predict differences in comparative SVE. We hypothesized that the SVE of honeybees would be lower compared to other bees, in crop systems, vertical plant tower and for plant species outside the native range of honeybees because previous studies have suggested such trends . Third, is there a correlation between floral visitation frequency and SVE? We evaluated this question separately for communities where honeybees were present or absent. We expected to find a positive correlation between visitation frequency and SVE that would be reduced when honeybees were present because honeybees are often highly frequent visitors and might be less consistently effective. Although previous studies have synthesized subsets of the pollination effectiveness literature , this paper is, at present, the most extensive meta-analysis to synthetize published results concerning single visit effectiveness.We performed a Web of Science search using a multiterm query designed to capture the highly variable terminology describing pollination effectiveness detailed in Ne’eman et al. . In May 2020, this search yielded 1,036 results. One of us screened the abstracts found by WoS to determine whether they potentially contained single visit effectiveness data. This yielded 388 papers. We also performed a Google Scholar search of the literature using a similar multi-term query , which yielded 116 additional papers. We found 62 papers from the reference sections of previously included papers. After removing duplicates and reading abstracts, we identified 468 papers which seemed appropriate for a more thorough screening. We followed the PRISMA protocol for collecting and screening data from the literature . To be included in our analysis, the paper had to contain empirical data on the per-visit contribution of at least one free-foraging visitor to plant reproduction. We considered pollen deposition, percent fruit set, fruit weight, and/or seed set as measures of SVE. Most studies were conducted with intact flowers, but we also included data from experiments that used the “interview stick” method . We did not include estimates of SVE based on equations or model outputs nor did we include data from trials that manipulated dead bees to deposit pollen. We extracted means, sample sizes, and measures of error directly from the text of the paper or from graphs using WebPlotDigitizer . When lower and upper error estimates were not symmetrical, we used the upper error estimate. When possible, we converted measures of error to standard deviation. When a paper did not report sample sizes, error, or other important information, we contacted the study authors. If we were unable to retrieve or estimate information on mean effectiveness and error, we excluded the paper from our analysis. We also excluded papers if we couldn’t convert other measures of error to standard deviation . After screening papers, 168 studies remained in our analytical dataset. We also extracted data on study year and location, plant species, plant family, whether the plant species was a crop-plant, pollinator taxon, pollinator group , and the native range of pollinator and plant species. We determined range status to bio-geographical realms by looking up the nativity of each taxon in the scientific literature and using occurrence records on the Global Biodiversity Information Facility website. If papers reported SVE outcomes from multiple sites or years, we extracted these data as separate outcomes and dealt with their non-independence statistically . We collected information on the visitation rates of pollinators if it was reported for the same plant species for which pollinator effectiveness data were reported. This rate could be reported as the number of visits to a focal flower or patch of flowers per unit time or the number of flowers visited per unit time and/or per unit area. We did not include data on the relative abundance of different visitors unless data were collected in a homogeneous landscape in which most visitors would have been visiting the focal plant species. If a study reported visitation data, we matched those data to the corresponding SVE data from the same study and plant species. Perfect matches required that pollinator taxa were reported to the same taxonomic resolution and that data were collected in the same year and location. When more than one measure of visit frequency was reported we preferentially used data on the number of visits to a focal flower per unit time. When more than one measure of SVE was reported, we preferentially chose whichever measure was better represented in our data, such that pollen deposition data were chosen over seed set data and seed set data were chosen over fruit set data.To address questions about the single visit effectiveness of honeybees and non-honeybees, we defined the effect size as the standardized mean difference of SVE values between honeybees and non-honeybees for each unique study, plant, site, and year combination. We chose to use Hedges’ g over other effect sizes because it is commonly used in the ecology literature for comparing two means , and it includes a correction for small sample sizes, which occurred with our data. Following Hung et al. , we calculated effect sizes for two separate comparisons: the difference between honeybees versus the most effective non-honeybee taxon and the average difference between honeybees and non-honeybee taxa . The SMD value is > 0 when other pollinators are more effective than honeybees and < 0 if the opposite occurs.