Defensive shimmering responses in Apis dorsata are triggered by dark stimuli moving against a bright background

Honeybee colonies consist of several thousand individuals and brood cells, as well as abundant stores of honey and pollen that must be guarded from predators (Fuchs and Tautz, 2011; Seeley, 1995). The giant honeybee Apis dorsata is a tropical open-nesting species with colonies occupying cliff faces, tall trees or artificial structures such as water tanks and window ledges (Misra et al., 2017; Neupane et al., 2013; Roy et al., 2011; Tan et al., 1997; Thomas et al., 2009). Honeybee species exhibit a wide array of anti-predatory behaviours (Fuchs and Tautz, 2011; Phiancharoen et al., 2011; Seeley et al., 1982), and the shimmering behaviour reported in giant honeybees is the first line of defence against aerial flying predators such as birds and hornets, which are their major predators (A. dorsata: Kastberger et al., 2011; Apis laboriosa: Batra, 1996). This behaviour is visualised as a dark travelling wave on the comb surface, and can precede a more aggressive collective stinging response (Seeley et al., 1982; Kastberger and Sharma, 2000; Kastberger et al., 2011; Wongsiri et al., 1997).

Shimmering in A. dorsata arises from ‘flickering’ behaviour, a form of dorsoventral abdomen flipping that occurs stochastically (Weihmann et al., 2012). Once initiated at trigger centres on the bee curtain (Kastberger et al., 2010a; Schmelzer and Kastberger, 2009), the shimmering wave follows the path of the flying intruder (Kastberger et al., 2010a). More trigger centres can subsequently form if the intruder persists, as quiescent bees from other regions on the comb initiate shimmering, leading to progressively stronger and larger displays (Weihmann et al., 2012). The waves also accelerate at a few trigger centres that detect approaching waves from further away (Kastberger et al., 2012, 2014). Shimmering confuses potential predators, and serves as a multi-modal alerting signal for nestmates using mechanoreceptive cues (Kastberger et al., 2011) or low-frequency comb vibrations (Kastberger et al., 2013), or by the release of Nasonov pheromone during wave generation (Kastberger et al., 2010b).

In this study, we used artificial moving stimuli that simulate aerial predators to examine the precise relationships between the visual features of the stimuli and the strength of shimmering, by varying the relative contrast and visual angles subtended by the stimuli on the bee's eye at different ambient light levels during daylight and twilight in this facultatively nocturnal honeybee (Dyer, 1985; Somanathan et al., 2009). We predicted that the strength of shimmering would increase under the following conditions: (1) with increasing contrast between the object and the background, (2) with larger stimuli and (3) at higher ambient light levels. We also examined whether repeated exposure to stimuli leads to a progressive reduction in the strength of shimmering through habituation over the course of the experiment. As shimmering occurs in response to potential threats from predators/intruders with consequences for colony survival, we expected that habituation would not occur. We thus predicted that the shimmering response would not reduce or dampen with recurring exposure to the stimuli over trials.

The experiments were conducted in October–December 2020 and in June 2022 using two natural Apis dorsata Fabricius 1793 colonies located on the Biological Sciences building at the IISER Thiruvananthapuram campus, India (8.68°N, 77.14°E). Both colonies consisted of a single comb and nested under the roof rafters. The first colony, henceforth referred to as hive A (1 m width, 1.2 m height), was situated 2 m above floor level on the open terrace of the building (Fig. 1A; Movie 1) and the second colony, hive B (1.8 m width, 0.7 m height), was 5-storeys above ground level (Fig. 1B; Movie 1). Although hives of A. dorsata are often found hanging from rooftop rafters and window ledges of multi-storey buildings, accessing hives for performing controlled experiments is often risky or challenging. Both experimental hives were naturally located in their respective positions. Translocating hives to convenient locations for ease of experimentation has been done in other studies (Young et al., 2021); however, as this could trigger absconding or change the baseline state of the colony, we refrained from doing so here.

The stimuli used in this study were made from circular cardboard cut-outs that varied in shade (light-grey, dark-grey or black) and size (diameter: 4, 8 or 16 cm; visual angle subtended on the eye of the bee: 1.6, 3.3 and 6.5 deg, respectively). The relative contrasts of the stimuli were further varied by using two different backgrounds (grey or black). In the case of hive A, a frame that consisted of two panels that could be swivelled along their vertical axes was placed 1.4 m from the colony (Fig. 1C), and the front and rear sides of the panels were covered with grey and black paper, respectively (paper procured locally). The rear side of each stimulus was attached to a string and moved either vertically or horizontally by the experimenter standing behind the frame and out of sight of the colony. To achieve this, the string was strung around the setup vertically or horizontally. Moving the strings without a stimulus was treated as the experimental control. The position of hive B on an overhanging roof rafter obstructed by walls and ledges prevented the use of the same arrangement. Instead, one of two backgrounds consisting of cardboard sheets (1 m×0.6 m) covered with black or grey paper was hung in front of the wall facing the colony during the experiments (Fig. 1D). The hive-to-background distance of 1.4 m was the same as for hive A. The experimenter stood behind the wall and moved the stimuli horizontally with the help of a wire, and moving the wire alone acted as an experimental control. Videos of the trials were recorded using a Sony Handycam HDR CX-405.

We carried out 799 trials on hive A, randomising the order of presentation of the stimuli across the following variables to prevent nestedness: stimulus shade, stimulus size, background colour, orientation of motion (vertical/horizontal) and side of the setup on which the trial was done (left/right). We conducted 240 trials on hive B, randomised across stimulus shade, stimulus size and background. Trials for both hives were carried out at least 15 min apart to ensure that the colonies reached a state of quiescence, and 9–45 trials were conducted each day between 06:00 h and 18:00 h (Table S1).

The effect of ambient light intensity on shimmering was examined by moving the largest black stimulus against the grey background under dim light conditions prevailing during astronomical, nautical and civil twilight (hive A n=26, hive B n=20; illuminance between 0.002 and 45 lx), and the responses were compared with shimmering elicited in bright daylight (illuminance >1000 lx). Illuminance levels were measured in lux using a Hagner Screenmaster (hive A) and a Hagner E4-X Digital Luxmeter (hive B) (Table S2).

To quantify whether habituation occurred as a result of repeated exposure within experimental days, we examined the strength of the shimmering response over trials. We excluded the first data point from each day as more than 12 h had elapsed after the last trial.

The light-grey, dark-grey and black stimuli had contrasts of 16.4, 6.6 and 0, respectively, against the black background, and 2.4, 0.5 and −0.8, respectively, against the grey background (Fig. S1B).

We compared the strength of shimmering (S_trial) in relation to stimulus shade, size and background. As the values of S_trial belong to the interval [0,1], interactions were modelled using beta regression with the hive identity as a random effect (Douma and Weedon, 2019). The side of the frame against which the stimuli were presented (left side/right side) and the direction of motion (vertical/horizontal) which were varied only in hive A were excluded as the shimmering response for these showed no significant differences in pairwise comparisons with Bonferroni correction (Fig. S2).

Shimmering responses were only observed in trials where the black stimulus was moved against the grey background (Fig. 2). The hierarchical model with hive identity as random effect and variable precision (φ) across stimulus size (model 4) was found to be the most parsimonious compared with the other models, including model 2 where hive identity was considered as a fixed effect (Table S4). Hence, we did not carry out post hoc statistical comparisons between hives. Shimmering in hive A was strongest with the largest black stimulus (subtending a visual angle of 6.5 deg, S_trial=0.25±0.12) and reduced in strength with the intermediate-sized black stimulus (3.3 deg, S_trial=0.09±0.10), whereas the two stimuli elicited an equally strong response in the case of hive B (6.5 deg S_trial=0.20±0.05, 3.3 deg S_trial=0.15±0.04). The smallest stimulus (1.6 deg) did not elicit shimmering in either hive (Fig. 2). Thus, the detection threshold for a moving dark object is probably between 1.6 and 3.3 deg. The distinct behavioural responses and the contrasts of the stimuli (Fig. S1) suggest that bees can distinguish and respond to the stimuli on the basis of contrast and size, leading to strong shimmering responses as the visual angle subtended by the intruding object/predator increases in the visual field of the bees.

In 31 trials out of the 45 carried out during dawn and dusk with the 16 cm black stimulus against the grey background, shimmering was not elicited (hive A n=13 and hive B n=18; illuminance between 0.1 and 3.9 lx), and the rest showed varying strengths of shimmering (illuminance between 3.9 and 45 lx). The model in which the data were pooled for the two hives was the most parsimonious (Table S5). The mean strength of shimmering under bright daylight conditions was an order of magnitude higher than the response during twilight hours (daylight S_trial=0.243±0.101, twilight S_trial=0.022±0.053; Fig. 3).

As A. dorsata shimmered only in response to the 8 and 16 cm black stimuli moving against the grey background, we excluded trials with other stimulus–background treatments from the analyses examining whether habituation occurred. Because the order of presentation of stimuli and backgrounds was randomised (see Materials and Methods), the black-on-grey trials were interspersed with other treatments and the trials were often not successive; 51 trials from hive A and 14 trials from hive B satisfied the condition of black stimuli (8/16 cm diameter) moving against a grey background and repeating within a day (06:00 h–06:00 h). We did not observe any significant change in the strength of the shimmering response with time elapsed between trials (range: 15–622 min; Fig. 4A), and there was no effect of time of day on the strength of the shimmering response (Fig. 4B).

Defence behaviour in honeybees involves several sensory modalities (Fuchs and Tautz, 2011), and shimmering is predominantly mediated by visual cues (Kastberger et al., 2010a, 2011). When we simulated predator movement using artificial stimuli, the strongest shimmering responses were obtained only in bright ambient light when the stimulus was much darker than the background. This is unlike the role of contrast in the context of foraging (in A. mellifera), where the stimulus can be either brighter or darker than the background and yet elicit similar responses (de Ibarra et al., 2000). We also found that the strength of shimmering decreased as stimulus size reduced and resulted in no response with the smallest stimulus. Although habituation or sensitisation occurs in some invertebrates in response to repeated exposure to a threat (Evans et al., 2019), we did not observe a reduction in the strength of the shimmering response for the inter-trial intervals employed in our study. While we did not find evidence for habituation, we do not exclude the possibility that more frequent exposure could lead to habituation and weaker shimmering responses.

Detection of dark objects moving against bright backgrounds (e.g. the sky) has been reported in several insects, including bees, although mostly in the context of mate detection or predation (Bergman et al., 2015; Sauseng et al., 2003). Male carpenter bees (Xylocopa tenuiscapa) detect and chase inanimate projectiles, mistaking them to be potential mates or rivals, when these objects subtend very small visual angles of 0.4–1 deg (which darken the visual field of a single ommatidium by as little as 2%) (Somanathan et al., 2017). Apis mellifera drones fly towards moving objects subtending as little as 0.4 deg (which darkens the visual field by just 8%) (Vallet and Coles, 1993). Apis mellifera workers can also detect a small black stimulus (0.6 deg×0.6 deg) moving at 65 deg s⁻¹ against a white background (Rigosi et al., 2017), which is significantly smaller than their detection threshold during foraging (3.7–5 deg) (Giurfa and Vorobyev, 1998). Anatomical estimates from the compound eyes of A. dorsata suggest coarser visual resolution compared with that in other honeybees (Somanathan et al., 2009), which agrees with recent estimates of detection thresholds in A. dorsata in the context of foraging (achromatic: 5.7 deg, chromatic: 12.4 deg; G. S. Balamurali, E. J. Warrant and H. Somanathan, in prep.).

Our experiments suggest that A. dorsata shimmers in response to dark objects moving against a bright background that subtend a visual angle of at least 3.3 deg (and probably even smaller, between 1.6 and 3.3 deg), which is smaller than their achromatic detection threshold while foraging. Using a conservative detection threshold of 3.3 deg for shimmering behaviour, and assuming that a colony shimmers in response to a hornet (a common predator) which is ∼2.3–2.7 cm in length (Girish and Srinivasan, 2010), the detection distance threshold corresponds to ∼40–47 cm, which agrees with results from real-world interactions involving free-flying hornets near A. dorsata colonies (Kastberger et al., 2008). Besides predators, A. dorsata also shimmer in response to intruders such as non-nestmate bees which display erratic flight patterns and splay their legs outwards while doing so (Weihmann et al., 2014). Based on a length of ∼2 cm for an A. dorsata worker (S.V., E.J.W. and H.S., personal observation) and a detection threshold of 3.3 deg, we speculate that the distance at which shimmering is triggered is around 35 cm. This suggests that shimmering could be a general response to aerial threats.

Open-nesting honeybees such as A. dorsata are likely to perceive flying predators as dark objects moving against the bright sky. Hornets are major predators of A. dorsata hives (Kastberger et al., 2013; Seeley et al., 1982), and Vespa tropica (the greater banded hornet) is a common predator in our study location (Balamurali et al., 2021). Moreover, hornets can act as strong evolutionary drivers in the evolution of specialised anti-predatory responses in honeybees (Cappa et al., 2021; Ken et al., 2005; Mattila et al., 2020, 2021; Papachristoforou et al., 2007; Tan et al., 2013). Shimmering resembles the anti-hornet ‘I-see-you’ display found in the sympatric cavity-nesting A. cerana in which guard bees at the entrance shake their abdomens in a to-and-fro motion in the presence of hornets, which supposedly warns the predator that the bees have detected its presence (Tan et al., 2013). Simulating this threat with black discs resulted in strong abdomen-shaking responses in A. cerana to stimuli of size 8 deg moving at 140 deg s⁻¹ (Koeniger and Fuchs, 1973, as cited in Fuchs and Tautz, 2011), which agrees with the achromatic visual detection threshold of 7.7 deg in A. cerana (Meena et al., 2021). In the context of an anti-predatory response such as shimmering, habituation may prove costly, and we found that the shimmering response does not reduce in strength after repeated exposure to the stimuli, which can be advantageous to ward off repeated approaches/attacks by predators.

Although in our study we found that the shimmering response increased with larger stimulus size, up to a maximum target size of 6.5 deg used in our experiments, we cannot rule out the possibility that the response increases further for even larger targets, or simply saturates. Beyond a certain size threshold [which is between 6.5 deg (current study) and 9.5 deg (Koeniger et al., 2017)], the colony responds to the object with stinging attacks (Koeniger et al., 2017), which resembles the defensive response of A. dorsata against much larger predators such as birds (Kastberger and Sharma, 2000). The findings from our study using moving objects are consistent with the notion that the shimmering behaviour of the giant honeybee is a response to an aerial predator moving in front of the hive against a bright sky background.

We thank G. S. Balamurali for his valuable inputs, and P. C. Anand for helping with the pilot experiments for this study. We also thank three anonymous reviewers for their valuable comments on the manuscript.