To characterize antennal response to ambient airflow, we provided flying bees with frontal airflow cues as would be experienced by them during forward flight. We then calculated the inter-antennal angle (IAA), defined as the angle between the lines joining the base and tip of each antenna (Sane et al., 2007), as a measure of antennal position (Figure 1A). When the antennae move backwards, the IAA increases and when the antennae move forward, the IAA decreases. Of the three conventional variables used in such experiments (David, 1982), we experimentally set the windspeed (i.e. velocity of ambient airflow relative to ground) in a calibrated, laminar wind tunnel, whereas the bees controlled their own airspeed (i.e. velocity of body relative to ambient air) and therefore also their groundspeed (i.e. velocity of the body relative to ground), which is the vector sum of airspeed and windspeed (Figure 1B). As tethered bees are stationary relative to the ground, their airspeed equals windspeed. However, when freely flying in air currents, airspeed and windspeed are independent of each other. Analogous to swimmers swimming within water currents, insects control their airspeed relative to the air pocket drifting at some windspeed.

Figure 1 with 1 supplement see all

Download asset Open asset

Honeybees modulated their antennal position as a function of their absolute airspeed. When presented with fixed airflow but no optic flow, both tethered and freely-flying bees held their antennae at constant IAA throughout a flight bout (e.g. Figure 1—figure supplement 1A). As frontal airflow increased, both antennae moved forward and mean IAA at each airflow value decreased as a sigmoidal function of airspeed (Figure 1C,D). The linear region of the sigmoid lay between 1.5 to 3 m/s (grey bar, Figure 1C,D), changing by approximately 35^o (Figure 1—figure supplement 1B) in tethered and 25° (Figure 1—figure supplement 1D) in freely-flying bees, significantly decreasing with each step change in airspeed (step change = 0.5 m/s; *p<0.0001; Moore’s paired test; N=10). This behaviour was similar regardless of whether airspeed stimulus was presented in linearly increasing or decreasing steps (Figure 1C, Figure 1—figure supplement 1B) or in random order (Figure 1E, Figure 1—figure supplement 1C). Thus, antennal position is calibrated against absolute airspeeds, independent of time history. At airspeeds less than 1.5 m/s or greater than 3 m/s, mean IAA at each airflow plateaued in both tethered (Figure 1—figure supplement 1B,C) and freely-flying bees (Figure 1—figure supplement 1D).

Previous studies on antennal responses of tethered bees to changing airflow did not report a zone of saturation (Heran, 1957) perhaps due to coarser sampling of their data (Figure 1—figure supplement 1E). We, however, consistently observed sigmoidal mean IAA responses, similar to locusts (Gewecke, 1974). As frontal airflow in the wind tunnel increased, freely-flying honeybees modulated their flight to maintain roughly constant groundspeed of ca. 0.4 m/s (Figure 1—figure supplement 1F; also [Barron and Srinivasan, 2006]). In these bees, we observed no correlation between groundspeed and IAA (Figure 1—figure supplement 1G).

Previous researchers have implicated the antennal mechanosensory Johnston’s organs (JO) in sensing airflow cues (Gewecke, 1974; Heran, 1957; Yorozu et al., 2009). JO spans the pedicel-flagellar joint in the antennae of all Neopteran insects, and tracks the motion of the flagella relative to pedicel. In most insects, the JO consists of several hundred scolopidial units that are range-fractionated. These enable both exquisite sensitivity and narrowly-tuned sensing over a large range of stimulus frequencies (Yorozu et al., 2009; Dieudonné et al., 2014). Does JO also mediate the observed antennal response to changes in airspeed? To test this hypothesis, we attenuated JO feedback by gluing the pedicel-flagellar joint (see Materials and methods) in tethered honeybees, and measured their antennal responses to airflow. To control for the extraneous effects of glue on the antenna (e.g. due to added weight), we glued the second annulus of the flagella from the tip (sham-treatment) in a separate group of the bees. Application of glue at the JO does not affect the movement of the antennae because the pedicellar-flagellar joints have no muscles, and hence motion around these joints are passive. Unlike sham-treated bees which showed the typical sigmoid response to frontal airflow (Figure 1F, non-normalized data in Figure 1—figure supplement 1H, N=5), bees with restricted JO positioned their antennae at the onset of flight (Figure 1—figure supplement 1I; also [Krishnan et al., 2012]) but their mean IAA at each airflow was insensitive to changing airspeeds (Figure 1G, N=7; p>0.1, Moore’s paired test). These experiments established that JO input is required for antennal positioning during flight, but not at flight onset (Krishnan et al., 2012). Thus, the antennal positioning at flight onset is a separate process from the inflight modulation of IAA.

Flying insects rely heavily on optic flow for flight control, most notably during slower flight manoeuvres such as landing or hovering (Baird et al., 2013). For example, visual feedback is critical for regulation of groundspeed and height, or in centring trajectories through narrow corridors in diverse flying insects (bees: [Srinivasan et al., 1996; Baird et al., 2005], Drosophila: [Straw et al., 2010], moths: [Kuenen and Baker, 1982], review: [Collett et al., 1993]). It also influences antennal position in insects that antennate during walking (Ye et al., 2003; Honegger, 1981).

Under natural flight conditions, front-to-back optic flow stimulus accompanies frontal airflow stimulus. Does the antennomotor system also respond to optic flow in addition to airflow? To address this question, we simulated forward flight conditions by presenting tethered flying bees with changing temporal frequency of front-to-back optic flow cues on two LED screens, and measured their IAA (Figure 2A). Note that, previous studies on optic flow dependent behaviours in honeybees have typically demonstrated that they extract angular velocity (ratio of temporal and spatial frequency) cues from the image motion, independent of spatial (Srinivasan et al., 1991) or temporal frequencies (Baird et al., 2013). In the experiments reported here, we have kept spatial frequency constant and only varied temporal frequency as the main experimental variable which means that the angular velocity is the temporal frequency times some multiplication factor.

Figure 2 with 2 supplements see all

Download asset Open asset

In a single flight bout (defined as an uninterrupted session between initiation and cessation of flight), tethered bees first experienced a static screen of black-and-white gratings (i.e. temporal frequency = 0) followed by a sequence of gratings moving from 0 to 25 cycle/s (cps) in discrete steps of 1 cps to simulate increasing flight speed (red curve, Figure 2B; Figure 2—figure supplement 1A,E; summary figure showing data from all individuals, Figure 2F). In the subsequent flight bout, the temporal frequency decreased stepwise from 25 to 0 cps (green curve, Figure 2B; Figure 2—figure supplement 1A,E; summary figure showing data from all individuals, Figure 2G) to simulate decreasing flight speed. Each stimulus step lasted for 1 s and the entire protocol for 26 s. When temporal frequency increased from 0 to 25 cps, an initial sharp adjustment in IAA (red curve, Figure 2B, summary figure Figure 2F) was followed by a graded response between 0 to approx. 10 cps (grey bar, Figure 2B), beyond which the value reached a plateau. The response curve for linearly decreasing optic flow (25 to 0 cps, green curve, Figure 2B, summary figure Figure 2G) was similar to the increasing cues, except for the lack of initial sharp IAA adjustment from zero to non-zero optic flow (also Figure 2—figure supplement 1A,E).

In these experiments, the optic flow was patterned to monotonically increase or decrease. Graded IAA responses to such optic flow patterns suggested two possibilities.

First, as in the case of airflow-based response, IAA response curve is innately calibrated against specific temporal frequency values, independent of its time history. If true, the response curve should remain invariant when the temporal frequency of the stimulus is presented in random order. This is an unlikely possibility because absolute values of optic flow are meaningless without prior knowledge of the spatial structure of the world.

Second, that bees respond to changes in temporal frequency rather than their absolute value. If true, IAA would increase (or decrease) in response to positive (or negative) changes in temporal frequency within the operating range (e.g. 0–10 cps for the bee in Figure 2B–E) while remaining unchanged outside this range (i.e. >10 cps). The threshold value was calculated separately for each bee (see Materials and methods for details). The latter criteria lead to specific predictions of when to expect changes in IAA (filled black circle; Figure 2C,D) and in which direction (arrow; Figure 2C,D).

To test the above hypotheses, we presented tethered honeybees with grating speeds between 1 to 25 cps in randomized order (Figure 2C–E and Figure 2—figure supplement 1B–D, Figure 2—figure supplement 1F–H) and compared the measured vs. predicted IAA response (Figure 2D; also Figure 2—figure supplement 1C,G). For each bee, we scored these as the fraction of correct predictions; a score of 1 corresponds to all correct, and 0 to all incorrect predictions. In all the cases, the total prediction score matched the actual observed score well above chance levels (*p<0.001; Student’s T test; see Materials and methods for details). When stimulus bins were rearranged in ascending order of temporal frequency, the reshuffled mean IAA response was not graded (compare Figure 2B and E; Figure 2—figure supplement 1A and D; Figure 2—figure supplement 1E and H) ruling out the possibility of innate calibration of IAA against optic flow. Thus, our data show that antennomotor activity in honey bees tracks changes in temporal frequency, and not their absolute value.

Bees typically experience the steepest gradients in optic flow during landing or hovering over flowers when optic flow rates are low (Baird et al., 2013). Hence, in the follow-up experiments, we focused on the lower range of stimuli from 0–1.8 cps (Figure 2H–J). Even in this narrow range, IAA responses tracked the magnitude and direction of grating patterns on the screen. For example, bee antennae robustly tracked simple sinusoidal stimuli that were both, cycled between front-to-back (0 to 1.8 cps), and back-to-front grating movement (0 to -1.8 to 0 cps) over the 10 s duration (Figure 2H), thus verifying that the antennae move in both directions as a function of optic flow. To characterize their responses in this narrow range of temporal frequency (0–1.8 cps), the tethered bees initiated wing flapping in front of a blank screen, followed by a sequence of black-and-white gratings moving front-to-back from 1.8 to 0.3 cps. Each grating sequence lasted for 6 s, interspersed with a static screen (zero optic flow) for 3 s. Despite inter-animal variability in the set points for IAA at zero optic flow (Figure 2—figure supplement 2A), all bees increased their IAA as a function of optic flow (Figure 2I), with each step change of 0.3 cps eliciting significant increase in mean IAA (*p<0.0001; Moore’s paired test, N=10) at that optic flow rate. The IAA response of tethered bees to optic flow remained intact (Figure 2J; Figure 2—figure supplement 2B) even when their JO were restricted by gluing the pedicel-flagellar joint, establishing that it is independent of the JO pathway. Thus, front-to-back optic flow causes antenna to move backward, opposite to frontal airflow which causes antenna to move actively forward.

The experiments outlined above established that inputs from two sensory modalities independently helped the bees to sense airspeed and optic flow. Whereas the JO-based antennal mechanosensory feedback was calibrated against absolute values of airspeed, visual feedback was calibrated relative to changes in optic flow under or across a threshold value. These experiments required flying bees to respond to single sensory cues at a time. Hence, we next measured IAA response of tethered bees to simultaneous visual and airflow cues. The paired stimuli combinations were arbitrarily drawn from values in previous experiments, and are not naturally correlated.

How do multimodal cues affect IAA response when simultaneously presented? Optic flow (red, Figure 3A) and airflow (blue, Figure 3A) cues were presented in three combinations, represented by three regimes A, B or C (grey bars; Figure 3A). Responses of a single bee are shown in Figure 3B,C. In each trial, we started the experiment with an intermediate bimodal combination (grey bars, middle Regime B; optic flow=0.9 cps, airspeed =2.5 m/s) and in random order, either decreased (to Regime A; optic flow =0.3 cps, airspeed =0.5 m/s; Figure 3B) or increased these values (to Regime C, optic flow =1.8 cps, airspeed =4 m/s; Figure 3C). Antennae maintained position when the cue combination changed from Regime B to Regime A (green line; Figure 3B; Pp>0.1, Moore’s test), but not when only airflow or optic flow was changed. When airspeed alone (i.e. with static grating) decreased from 2.5 to 0.5 m/s (blue line; Figure 3B; p<0.001, Moore’s test), both antennae moved backward. However, they moved forward when optic flow alone (i.e. with fan off) decreased from 0.9 to 0.3 cps (red line; Figure 3B; p<0.001; Moore’s test). Similarly, when the cue combination changed from Regime B to Regime C, antennae again maintained position (green line; Figure 3C; p>0.1, Moore’s test), but moved forward when airflow alone increased from 2.5 to 4 m/s (blue line; Figure 3C) and backward when optic flow alone increased from 0.9 to 1.8 cps (red line; Figure 3C). Thus, visual and airflow cues elicit opposite IAA responses which, when acting in concert, maintain antennal position.

Figure 3

Download asset Open asset

This was also borne out in the pooled data over multiple trials (Figure 3D,E). Decrease of only airspeed resulted in a negative mean ΔIAA (blue, Figure 3D; *p<0.001; Moore’s test), whereas increase resulted in a positive mean ΔIAA (blue, Figure 3E; *p<0.001; Moore’s test). Similarly, a decrease in only optic flow resulted in positive mean ΔIAA values, but its increase led to negative mean ΔIAA (red, compare Figure 3D,E; *p<0.001; Moore’s test) compared to the IAA values in Regime B. Here, mean ΔIAA values significantly differed from a hypothetical mean of zero (one-way ANOVA with post-hoc Tukey’s Honest Significant Difference test, *p<0.05). However, when both cues were co-varied, ΔIAA did not significantly change (green, Figure 3D,E; p>0.1, Moore’s test) and their means were statistically indistinguishable from zero.

How does the presence of one cue alter the antennal response curve of the other cue? We considered three possibilities. First, if the response curve of Cue A remains unaltered in presence of a constant Cue B, then it is likely that the system adapts to cue B. Second, if the response curve to Cue A is uniformly offset by the presence of constant Cue B, then the cross-modal influence is likely linear and summative. Third, if response curve to Cue A is non-uniformly offset in presence of constant Cue B, then crossmodal influences are likely to be non-linear. Testing for possibilities provides insights into how inputs from different modalities combine to determine the antennal position.

We measured IAA against optic flow in presence of two constant values (0 and 1 m/s) of airspeed. The two curves were uniformly offset over the range from 0 to 1.8 cps (compare dotted and solid lines, Figure 4A,B; *p<0.05, Moore’s test, N=8). Similarly, for airspeeds between 0 and 4 m/s, we observed a steady offset for greater optic flow, but in the opposite direction (compare dotted and solid lines, Figure 4C,D; *p<0.05, Moore’s test, N=9). In both experiments, the presence of constant optic flow (or airflow) altered the set point of the antenna to a new mean value that is greater (or less) than the original set point (Figure 4B,D) and the antenna responded to variation in airflow (or optic flow) relative to the new set point. Thus, the antennomotor system linearly combines the multimodal cues and recalibrates accordingly.

Figure 4

Download asset Open asset

The above study provides us with insights into how the insect antennomotor system integrates the multiple sensory cues that it encounters during flight. This can be summarized in a general schematic of the multi-modal integration of antennal positioning behaviour in flying honeybees (Figure 4E). For the sake of completeness, this model must also include the role of the mechanosensory hair plates (or Böhm’s bristles) in reflexive antennal positioning, which appears conserved across insects according to our studies in moths (Krishnan et al., 2012), bees and crickets (Sant and Sane, 2016). To summarize these studies: antennal hair plates are stimulated when the antenna undergoes a substantial movement in the scape-head capsule or pedicel-scape joints. The mechanosensory neurons underlying these bristles directly project into the Antennal Motor and Mechanosensory Center (AMMC) where they arborize on the dendritic fields of antennal motor neurons and activate them. Any gross change in antennal position thus elicits a rapid reflexive correction of the antennal position with latencies under 10 ms in hawk moths (Krishnan et al., 2012) and probably on the same order in honeybees and other insects. Based on these data, we proposed a model to describe the antennal positioning reflex loop mediated via antennal mechanosensors (Krishnan et al., 2012). Although arrangement of the hair plates in different insects varies, the underlying neural circuitry and hair plate function is conserved in moths, bees, and most other insect orders (Krishnan and Sane, 2015). Previous studies in hawk moths have described only how the antenna, once positioned, reflexively maintains this position (Krishnan et al., 2012), and also their response to visual cues (Krishnan and Sane, 2014). However, the combinatorial role of these cues in antennal positioning behaviour during flight remained unclear.

Here, we show that the reflexive maintenance of in-flight antennal positioning is simultaneously modulated by both visual (from eyes) and airflow (from JO) feedback during flight. This study also shows that their action is via multi-modal pathways which combine in a mutually antagonistic fashion; whereas frontal airflow detected by JO reduces IAA, the front-to-back optic flow detected by compound eyes increases IAA. The combination of these inputs modifies IAA to a new set point, which is then maintained by the antennal hair plates via a negative feedback loop that ensures rapid maintenance and correction of the antennal position. The block diagram in Figure 4E thus provides a model for multimodal sensory control of antennal positioning response. It is possible that other inputs, such as olfaction, also additionally modulate the antennomotor responses, which is the subject of future studies.

Our study shows that the initial positioning of antennae at the flight onset (antennal deployment) is separate from its later inflight maintenance by the multimodal inputs (inflight antennal positioning). Both behaviours require the hair plate-mediated reflex pathway but the set-point of this reflex system is under multimodal control in the latter case. In particular, restriction of the JO feedback has no effect on the antennal deployment behaviour, whereas it completely disrupts inflight antennal positioning. It has been suggested that inflight modulation of IAA maintains the scolopidial units of the JO in their operating range (Hinterwirth et al., 2012). Although this hypothesis remains unaddressed and is beyond the scope of this paper, our data clearly show that inflight antennal movements are precisely modulated, and product of both visual (from compound eyes) and mechanosensory (from the hair plates and JO) input. Visual feedback also induces directionally-sensitive antennal movements in other insects such as hawk moths (Krishnan and Sane, 2014), Drosophila (Mamiya et al., 2011), and many orthopteran insects (Honegger, 1981; Ye et al., 2003). Unlike the hair plate reflexes which are strictly unilateral (Krishnan et al., 2012), visual feedback drives the activity of both ipsi- and contralateral antennal motor neurons in moths (Krishnan and Sane, 2014) and may therefore serve to coordinate the movements of both antennae. The mechanosensory pathways mediated by the hair plates then provide rapid reflexive, local correction of the intended position of each antenna.

Recent studies have also described how the integration of antennal mechanosensory and visual inputs mediates abdominal flexion in moths (Hinterwirth and Daniel, 2010) which in turn is relevant for flight control and balance. In honeybees too, their combined effect on the abdominal streamlining behaviour enables drag reduction during flight (Luu et al., 2011; Taylor et al., 2013). In freely-flying moths, the stimulation of antennal muscles elicits abdominal flexion accompanied by a change in flight trajectory (Hinterwirth et al., 2012). Similarly, integration of mechanosensory and visual information mediates flight control in other insects such as Drosophila (Sherman and Dickinson, 2003; Fuller et al., 2014). Whereas the relative influence of mechanosensory input is greater during rapid turns, visual input is more important during slower rotations (Sherman and Dickinson, 2003; 2004).

The integration of these multimodal cues quite likely occurs in the Antennal Motor and Mechanosensory Centre (AMMC) region of the insect brain, which houses the soma of the antennal motor neurons. This region also receives the arbors of motion-sensitive visual interneurons (Hertel and Maronde, 1987), and likely also the inputs from cephalic hair mechano-afferents. These different modalities are known to influence diverse flight behaviours. It is thus likely that descending interneurons transduce multimodal sensory information that elicits these behaviours. In locusts, the tritocerebral commissure giant (TCG) interneurone integrates visual and airflow information (Bacon and Tyrer, 1978), while other descending interneurons in crickets integrate antennal mechanosensory and visual inputs (Gebhardt and Honegger, 2001). The importance of AMMC as the site of integration of multimodal sensory feedback is an important topic of future study.

The data on antennal position are also relevant to studies on speedometry in honeybees. A key finding of this paper is that IAA responds to absolute airspeed values from 1.5 to 3 m/s, which is narrow compared to the typical range of airspeeds (between 0 and 7 m/s; [Wenner, 1963])of honeybees in free flight. IAA is also calibrated against optic flow cues, but this feedback depends on the spatial structure of the environment. As shown in studies over the past two decades, the odometer in honeybees is quite clearly visual (Srinivasan, 2014). The visual system can extract the overall image motion, which is perceived by and integrated over the motion detectors in the retina (Zanker et al., 1999). Honeybees trained to find food at a specific distance in tunnels lined with visual stripes can correctly judge distances even when the spatial frequency of the stripes is altered (Baird et al., 2005; Srinivasan et al., 1996; 1997; 2000; 2014). Moreover, they estimate distances correctly despite head or tailwinds, suggesting that parameters such as windspeed, flight duration and number of wing beats etc. play no role in the odometry. Indeed, image motion cues can also be used to estimate depth and distinguish between near and far objects independent of their size (Kirchner and Srinivasan, 1989; Srinivasan et al., 1989; Zhang et al., 1992). Thus, bees appear to be able to separately process size and velocity cues (Srinivasan et al., 1993).

Complement to the visual odometer, we show here that honeybees are also able to extract information about both airspeed and groundspeed. Importantly, the airspeed calibration mediated by JO is absolute, but works over a narrow speed range, whereas the groundspeed calibration mediated by optic flow can be tuned over a broader range. We propose the hypothesis that honeybees calibrate their optic flow speedometer using their JO-based system that measures absolute airspeeds (Figure 4F). At the onset of a regular flight bout, bees simultaneously sample the optic flow (red lines) and airspeed (blue line). The three red lines correspond to temporal frequency of optic flow that is slow (i.e. if objects are far away) to fast (i.e. if objects are close). Sampling the optic flow (Figure 4F; black segments on the red lines) corresponding to the dynamic range of the JO-based airspeed sensor (e.g. between ~1.5 to 3 m/s) provides bees with the means to cross-calibrate average optic flow against absolute airspeeds. As the airspeed-based estimation only works in a narrow range from 1.5 to 3 m/s, the optic flow-based speedometer requires extrapolation well beyond the range of the antennal system. As shown in Figure 4A–D, the presence of one cue linearly offsets the response curve for the other, analogous to using a standardized albeit narrow-ranging measure (e.g. airflow-based speedometer) to calibrate another arbitrarily scalable measure (e.g. the optic flow based speedometer). A disadvantage of such a system is that it must be recalibrated at the beginning of each bout, or within bouts in case of sudden gusts of wind which may confound the calibration. Under such circumstances, we predict that honeybees would have to slow down to recalibrate their optic flow based speedometer against the airspeed-based speedometer, before resuming flight.

Our data throw some light on puzzling observations from previous studies. In their study on visual regulation of groundspeeds in honeybees, Barron and Srinivasan (Barron and Srinivasan, 2006) observed that honeybees freely flying within a tunnel lined with checkerboard patterns flew at speeds of approximately 0.4 m/s. However, when the checkerboard pattern was replaced by an axial-stripe pattern which offered sparse optic flow cues, its groundspeed was maintained with low variance at a value of 1.4 m/s. Maintenance of ground speeds (which equals airspeed in still air) requires that bees be able to ‘sense’ their speed even when optic flow cues are sparse. How is this possible if speed is maintained only via optic flow? To explain this result, we propose that, in absence of optic flow, the bees rely on their JO-based airflow sensors to both set and maintain groundspeed. It is worth noting that the value of 1.4 m/s lies at the cusp of the airflow response curve (Figure 1C), and is maintained as the new reference at different wind speeds in sparse optic flow (see Figure 3 in Barron and Srinivasan [2006]).

A JO-based antennal airflow sensing, can in principle offer a ‘true’ airspeed measure for odometry. It is not presently clear if the visual odometer would need such information as the current model for visual odometry is able to quite robustly explain most observations. One would expect such a speedometer system to accumulate errors as the range increases, as has also been observed with the honeybee odometer, which follows the Weber’s law (Cheng et al., 1999). Future studies will be required to determine if the several predictions that emerge from this model about honeybee speed control hold true. The antennal positioning responses thus provide unique insights into how honeybees and perhaps other insects sense and combine information from multiple modalities to help reduce ambiguities arising from drift in any one modality.

After capturing individual forager bees at the hive, we cold-anaesthetized them on ice until they were inactive. The anaesthetized bees were then dorsally tethered to a bent metal rod (30 mm in length and 0.2 mm diameter) using a synthetic rubber based adhesive (Fevibond, Pidilite, Mumbai, India). The bee was then provided with sucrose solution and left to recover for 45 min. When performing experiments involving visual stimuli, we dark-adapted the bees in a darkened box during this recovery period. To elicit flight, we provided the bee with a piece of tissue paper to hold, and then suddenly withdrew it to elicit flight due to tarsal reflex.

We constructed an open-circuit wind tunnel (30 cm x 30 cm cross-section and 120 cm in length, with a 40 cm x 30 cm x 30 cm test section) with a fan, driven by a motor drive at one end that drew air through the wind tunnel. The wind tunnel was calibrated using a constant temperature mini-anemometer (Kurz 490S, Kurz Instruments, Inc., Monterey, CA), which was modified to take direct voltage readings (Sane and Jacobson, 2006). The voltage readings were consistently reproducible between 0–6 m/s at a step size of 0.5 m/s. To minimise confounding visual cues, the floor and walls of the wind tunnel were covered with white paper.

To determine how antennal position varies with ambient airflow, we provided tethered and freely flying bees with frontal airflow between 0 and 5.5 m/s and between 0 and 3.5 m/s in steps of 0.5 m/s respectively. At their fastest, freely flying honey bees have been estimated to fly at ~7.5 m/s in natural outdoor environments. Presence of headwinds or tailwinds of about 3 m/s does not affect the speed of the bees (Wenner, 1963) in their natural flight. In both cases, we filmed bees from the top and side with two synchronised high-speed cameras and calculated their IAA. We performed both tethered and free flight assays in the wind tunnel to look at how the antennae respond to changes in airflow.

To investigate how visual motion influences the antennal positioning response, we modified a MATLAB demo routine from the Psychophysics toolbox (Mathworks Inc, Natick, MA) (Brainard, 1997; Kleiner and Pelli, 2007) to generate a moving visual grating pattern. The patterns were displayed on two LED monitors (Beetel 8”x14”, 640x480 resolution, 60 Hz refresh rate) and controlled simultaneously by a master computer equipped with a dual-VGA graphics card (nVIDIA GeForce 9800GT). These screens did not show noticeable flicker at 1000 fps, as determined by filming them with a high-speed camera. The monitors were placed in a V configuration, with a 2 cm gap between them. The tethered bee was positioned at the centre of the two monitors, facing the apex, at a linear distance of 10 cm from each monitor (Figure 2A). The floor below the screens was layered with white paper to increase the contrast of the antennae during filming.

We provided open-loop forward translational stimuli (simulating front-to-back movement of the visual field) to the tethered bees, at a spatial frequency of 0.44 cm^-1 and various temporal frequencies (0 to 25 cps for the first set of the experiments and 1.8, 1.5, 1.2, 0.9 and 0.3 cps for the second set of experiments), corresponding approximately to angular speeds of 0 to 300 deg/s and 18, 15, 12, 9, 6 and 3 deg/s respectively. Using two synchronized high-speed video cameras (Phantom v7.3, Vision Research, Inc. Wayne, NJ), we filmed tethered flying bees at 250 fps, from both a side view and an overhead view. A third camera, synchronized to the first two, filmed the visual stimulus being displayed to the bee. We provided each bee with a range of temporal frequencies over a single flight bout, and digitized and analysed the whole video, the entire flight bout of the bee in that trial for the first set of experiments. In the second set of experiments in which we gave the bees optic flow from 0 to 1.8 cps, we digitised and analysed 50 frames of video per temporal frequency for 50 wing beats. In our analysis, we discarded bees that did not fly continuously through the entire protocol.

To provide both visual and airflow stimuli to the tethered bees, we placed a 4 wire DC fan in front of the bee in a 2 cm gap between the two screens in the set up described above, such that the tethered bee received frontal airflow (Figure 2A). The airflow was collimated and fairly laminar, and we ensured that there were no other sources of airflow. The speed of the fan was adjustable and we could measure the speed of the frontal airflow near the head of the bee using a hot wire anemometer. The fan was placed inside a rectangular box with a plexiglass tunnel in front of it that opened between the two screens so that the tethered bee received collimated frontal airflow.

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Animal experimentation: Study involved experiments on honeybees and were conducted according to ethical guidelines.

© 2016, Roy Khurana et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.