The study was carried out in a blueberry experimental field located at the Agrarian Field Station of Viseu, Portugal (40°39^′52.4^′′ N, 7°54^′12.6^′′ W). Nineteen blueberry cultivars were evaluated, including 10 northern highbush (“Aurora”, “Bluecrop”, “Bluegold”, “Chandler”, “Draper”, “Duke”, “Elliot”, “Huron”, “Legacy”, “Liberty”), eight southern highbush cultivars (“Biloxi”, “Camelia”, “Misty”, “O'Neil”, “Rebel”, “Sharpblue”, “Star”, “Suziblue”) and one rabbiteye cultivar (“Ochlockonee”). Plants were cultivated in soil with rows oriented east–west, and there were 15–18 plants of a given cultivar within a row. The experimental field was originally set up to assess the suitability of these cultivars for growth and production under the region's climatic conditions. Northern highbush and southern highbush blueberry differ in the number of chilling hours required for normal floral development, with northern highbush having higher chilling requirements (800 to 1000 h), southern highbush having lower chilling requirements (400 to 800 h) and rabbiteye requiring around 600 h (Retamales and Hancock, 2018). We randomly selected five plants per cultivar along the respective row to evaluate flower morphological traits and perform pollinator observations.

The experimental field was located in a landscape with multiple land-cover and land-use types, including farming systems (a mix of fruit orchards), grasslands, and forest and urban areas, similar to many production areas in Portugal. The study site was pollinator friendly (i.e. low disturbance, low inputs, and diverse floral and nesting resources). No beehives were added to the study site.

Corolla length, width and aperture diameter were measured following the methodology of Courcelles et al. (2013), as these traits can potentially affect pollinator behaviour. Measurements were collected from 6–10 flowers for each of the five plants selected per cultivar. Flowers were collected into microtubes with 70 % ethanol and measured in the laboratory with a digital calliper. Briefly, we measured the corolla length by taking the flower's total length from the floral tube's base to the corolla's opening, the corolla's width at the widest point and the corolla opening diameter.

Floral visitors were observed at each cultivar's flowering peak on sunny days (temperature range 13–25 °C) with favourable weather conditions for pollination, from 15 March to 20 April 2023. Floral visits were recorded at different times of the day from 9:00 AM to 5:00 PM (CET+1 h), totalling 50 h 17 min of surveillance. The monitoring of floral visitors followed standard protocols (Dafni et al., 2005). Briefly, on each monitoring day, flowering plants were selected haphazardly and observed for 5 min periods. As several cultivars were simultaneously in flower, the observer alternated observations among cultivars after each 5 min period. Both the selection of individual plants and the sequence of cultivars were random. For each selected plant in each cultivar, we recorded the number and identity of individual insects that interacted with the flowers, insect behaviour (i.e. legitimate visit and primary or secondary nectar robber), and the cultivar and number of flowers visited. The interaction with each flower was considered a legitimate visit when the insect inserted its proboscis and/or head into the corolla or robbing when the insect positioned itself over the corolla, creating holes in its base (primary robber), or collecting nectar from existing holes (secondary robber). We used photographs and targeted captures to identify the insect visitors. Additionally, the total number of open flowers per cultivar was also recorded.

For each selected plant per cultivar, the visitation rates were calculated as the ratio of legitimately visited to open flowers per 5 min period. Values are presented as percentages of the flowers visited. In addition to calculating overall visitation rates, we computed the percentage of flowers visited by each of the three most common floral visitors (Bombus terrestris, Linnaeus, 1758; Anthophora plumipes, Pallas, 1772; and Apis mellifera) and the percentage of flowers robbed. For pollinator traits, we retrieved average values of body and proboscis length for these three species from Cappellari et al. (2022) to characterise the relative differences among these visitors.

Flower morphological data and visitation rates were averaged per individual plant in each cultivar. To test for the effect of cultivar on flower morphology, visitation rates and robbing, we used general linear models with corolla length, corolla width, corolla opening diameter, visitation rates or percentage of flowers robbed as a dependent variable and cultivar as a fixed factor. We used principal component analysis to integrate the flower morphological traits, which are correlated to each other, using prcom, FactoMiner and facto extra (Kassambara and Mundt, 2020; Lê et al., 2008). The scores of the first PCA axis were extracted and used as a predictor to explore the effect of flower size on pollinator behaviour. The relationship between flower morphological traits and visitation rates by B. terrestris, An. plumipes and A. mellifera was assessed using generalised additive models (GAMs), with the visitation rate by each floral visitor as a dependent variable and the PCA first axis scores as a predictor (gam(visit_insect ∼ s(PC1), data = data, method = REML)). The mgcv package was used, and the basis dimensions and model fit were evaluated using gam.check. We used the ggplot package to build plots. Model fit was assessed using the diagnostic plots of the DHARMa package (Hartig, 2016). All analyses were performed using R software, version 4.3.0 (Core Development Team R, 2016).

The flower morphology of the studied blueberry cultivars at the study site differed significantly (Fig. 2, Supplement Fig. S1). Southern highbush cultivars tended to have smaller corollas than northern highbush cultivars, while the rabbiteye presented intermediate values (Fig. 2, Supplement Fig. S1). Corolla length ranged from 6.21 mm in “Star” to 11.85 mm in “Chandler”, while corolla width ranged from 4.18 mm in “Rebel” to 7.97 mm in “Draper”. Finally, the corolla opening ranged from 2.26 mm in “Rebel” to 5.13 mm in “Aurora” (Supplement Table S1 and Fig. S1).

A total of 13 insect species were observed interacting with the blueberry flowers (Table 1). Among these, three species accounted for 93.8 % of insect interactions with blueberry flowers, with the wild buff-tailed bumble bee (B. terrestris; proboscis 6.50 mm long) being the most frequent visitor (Supplement Fig. S2) with visitation rates at 4.71 %. The longer-tongued wild bee An. plumipes (proboscis 9.00 mm long) was the second most abundant floral visitor, followed by the shorter-tongued managed honeybee (A. mellifera; proboscis 5.00 mm long).

The percentage of flowers with legitimate visits differed significantly across cultivars (X18= 84.66, p < 0.0001; Fig. 3A). The lowest values were observed in “Chandler” (5.36 %), while the highest values were observed in “Aurora” (16.5 %). The percentage of flowers robbed also differed among cultivars (X18= 130.53, p < 0.0001; Fig. 3B), with “Chandler” showing the highest values (2.87 %).

The results from the generalised additive models testing the effect of flower size – represented by the weighted combination of corolla length, width and opening (PCA axis 1) – on the behaviour of the main visitors of blueberry flowers show that floral visitors did not visit cultivars equally; B. terrestris had more visits to cultivars with intermediate corolla sizes (Fig. 4A; Table 2), An. plumipes visited more flowers on cultivars with large corollas (Fig. 4B; Table 2) and A. mellifera tended to visit cultivars with small corollas (Fig. 4C; Table 2).

Some studies indicate that a diverse assemblage of insects contributes to blueberry pollination, with honeybees showing the most significant abundance, followed by bumblebees and solitary bees (e.g. Cortés-Rivas et al., 2023b; Eeraerts et al., 2023). This pattern was also observed by Isaacs and Kirk (2010) on large commercial farms, but not on small farms, where they found that wild bees were the primary pollinators, and A. mellifera was less common than wild bees. In our study, B. terrestris and An. plumipes were the main visitors, followed by A. mellifera, while other wild bees were observed interacting at very low frequencies. Consistent with our findings, Cane and Payne (1993) report the predominance of a few species visiting rabbiteye blueberries in the southeastern United States, including honeybees, bumblebees and the southeastern blueberry bee (Habropoda laboriosa). Our study site is more closely aligned with the characteristics of small farms described by Isaacs and Kirk (2010) and Cane and Payne (1993), and is likely to provide areas with suitable habitats for diverse and abundant floral resources and nesting sites, contributing to a greater abundance of wild bees. The low abundance of honeybees could be partially due to the absence of beehives at the study site, unlike the standard practice in commercial orchards, and to the proximity of an apple orchard, which overlapped in flowering with blueberry and may be more attractive to honeybees due to easier access to resources offered by the open flower morphology.

Flowers such as those of the blueberry, with morphological traits that condition access to pollen and rewards, will affect pollinator visitation and behaviour. Indeed, our data support this, as we found distinct relationships between the visitation rates of each of the three main floral visitors and flower size. Bombus terrestris visited more flowers on cultivars with intermediate flower sizes, An. plumipes visited more cultivars with larger flowers and A. mellifera tended to visit cultivars with smaller flowers. These differences may be linked to the relationship between proboscis size and flower morphology, which limits access to rewards and determines pollinator behaviour (Klumpers et al., 2019). For example, bumblebee species that differ in proboscis length exhibit differences in flower visitation across a range of flower-tube lengths (Inouye, 1980). This may be related to flower handling time and efficiency in nectar extraction, where pollinators with proboscides shorter than the corolla require more time to collect nectar compared to pollinators with proboscides that match corolla size (Klumpers et al., 2019). Higher handling efficiency implies a lower cost-benefit in foraging and higher fitness, leading pollinators to prefer flowers they can handle more efficiently (Balfour et al., 2013). The average proboscis lengths of the main pollinators of blueberry may explain the patterns observed in our study. Apis mellifera, with the shortest proboscis (5.00 mm) of the three species, visited cultivars with smaller flowers; An. plumipes, with the longest proboscis (9.00 mm), visited cultivars with larger flowers; and B. terrestris, with intermediate values (6.5 mm), visited cultivars with intermediate sizes. While bees with long proboscides can visit both flowers with long and short corollas, there is evidence indicating that the handling time and the number of flowers visited per unit of time are positively correlated to proboscis length (Balfour et al., 2013; Herrera, 1989; Inouye, 1980). For example, proboscis length and resource extraction time were negatively correlated in lavender, and when the length of the corolla was experimentally reduced, the handling time by A. mellifera decreased (Balfour et al., 2013). Results suggest that the three main visitors of blueberry in our study select flowers that best fit their morphology, allowing them to forage more efficiently, and show evidence of resource partitioning, where An. plumipes visits flowers less accessible to and/or less efficiently handled by other visitors with a shorter proboscis. Nevertheless, other factors that influence insect behaviour, such as nectar volume and sugar content (Klumpers et al., 2019), have not been assessed, and their contributions need to be considered in future studies.

We observed primary nectar robbing by the carpenter bee Xylocopa cantabrita Lepeletier, 1841 and secondary robbing by B. terrestris and A. mellifera. Similar behaviours were also reported by Cane and Payne (1993) and Sampson et al. (2004) for X. virginica Linnaeus, 1771 (absent in Europe) and A. mellifera, where the carpenter bee pierces the base of the corolla, and honeybees profit from the greater accessibility to the nectaries and engage in secondary nectar robbing. Nectar robbing can influence the patterns of nectar availability and consequently affects the attractiveness of the flower to pollinators, resulting in lower rates of legitimate visits and lower fruit production (Castro et al., 2008). Nevertheless, despite this behaviour, carpenter bees are reported to deposit pollen tetrads on the stigma on their flower visits, thereby contributing to blueberry pollination, particularly in cultivars with protruding stigmas (Sampson et al., 2004; Tucker et al., 2019).

Overall, our findings indicate that a variation in flower morphological traits across blueberry cultivars influences pollinator visitation rates, with proboscis length likely linked to differences in nectar foraging preferences among the evaluated taxa. Breeding programmes are encouraged to consider flower traits and their influence on pollination success in commercial settings. Likewise, growers should select cultivars that can be readily pollinated by the insect communities they rely on for pollination services, given that pollination success is a prerequisite for commercially viable yields and fruit quality. Moreover, promoting wild pollinators by maintaining or creating natural areas with flower and nesting resources increases the probability of having species with different proboscis sizes, ensuring legitimate visits to a wider range of cultivars, as well as species with the capacity for buzz pollination (e.g. Bombus spp., Anthophora spp.), which are more efficient at pollinating flowers with poricidal anthers (Buchmann and Hurley, 1978; De Luca et al., 2019), such as those of the blueberry. Conversely, if the honeybee is to be the primary pollinator, care should be taken to select cultivars with shorter and/or wider corollas, allowing the honeybees to reach nectar by inserting their heads inside the corolla, preferably with extruding pistils that allow contact with honeybee body parts carrying pollen (Hoffman et al., 2018).