One persistent question in animal navigation is how animals follow habitual routes between their home and a food source. Our current understanding of insect navigation suggests an interplay between visual memories, collision avoidance and path integration, the continuous integration of distance and direction travelled. However, these behavioural modules have to be continuously updated with instantaneous visual information. In order to alleviate this need, the insect could learn and replicate habitual movements (‘movement memories’) around objects (e.g. a bent trajectory around an object) to reach its destination. We investigated whether bumblebees, Bombus terrestris, learn and use movement memories en route to their home. Using a novel experimental paradigm, we habituated bumblebees to establish a habitual route in a flight tunnel containing ‘invisible’ obstacles. We then confronted them with conflicting cues leading to different choice directions depending on whether they rely on movement or visual memories. The results suggest that they use movement memories to navigate, but also rely on visual memories to solve conflicting situations. We investigated whether the observed behaviour was due to other guidance systems, such as path integration or optic flow-based flight control, and found that neither of these systems was sufficient to explain the behaviour.

When insects, such as bees, wasps or ants, exploit a food location, multiple guidance systems are at work and steer the animals along their habitual routes (Buatois and Lihoreau, 2016; Kohler and Wehner, 2005; Lipp et al., 2004; Woodgate et al., 2016). The underlying navigation systems are thought to rely on comparing current sensory information with information memorised previously. One possibility for how this might be accomplished is comparison of the current scenery with familiar views memorised along previous journeys (view-based system) and then following the most familiar direction (Baddeley et al., 2012; Zeil, 2012). Another possibility is comparison of the current orientation of the animal with the outcome of continuous integration of the direction and distance travelled (path integration system; Heinze et al., 2018; Pfeffer and Wittlinger, 2016; Wittlinger et al., 2006), providing a vector pointing towards the end of the route.

Recent experiments indicate that path integration and view-based guidance systems may work in parallel in steering the animals along their paths (Bregy et al., 2008; Collett, 2012; Legge et al., 2014; Wystrach et al., 2015, 2019). A framework for how these guidance systems might be integrated has been successful in explaining several navigation behaviours of ants and bees (Hoinville and Wehner, 2018). However, the interplay of these guidance systems requires the continuous update and weighting of each module to yield the direction in which the insect should move along the route. When the insect makes a curve around an obstacle, for example, it needs to frequently update where to go to replicate the turn. Instead of frequently updating the direction in which to move, insects may encode the movement itself that needs to be carried out at specific locations, in the same way as a trained skier going downhill in a slalom will memorise every turn to travel as fast as possible. Although insects are capable of memorising and replicating a sequence of actions when olfactory and visual cues are not present (Collett et al., 1993; Zhang et al., 1996, 2012; Macquart et al., 2008; Mirwan and Kevan, 2015), movement memories have not usually been considered as an additional mechanism for following routes.

We analysed here whether movement memories could play a role along with view-based guidance systems for animals travelling on habitual routes. We addressed this question by using a novel experimental paradigm and bumblebees, known for their efficient route navigation capabilities (Buatois and Lihoreau, 2016; Kohler and Wehner, 2005; Lipp et al., 2004; Woodgate et al., 2016), as experimental animals. We habituated bumblebees to follow a route between their hive and a food source in a flight tunnel. Along the way, they had to avoid collisions with ‘invisible’ objects by using visual directional cues carried by neighbouring objects, forming a sequence of decisions of right or left turns. To test the hypothesis that bumblebees established movement memories during this habituation phase, we rearranged the visual directional cues in various ways so that visual information and the sequence of movement (a succession of left and right turns around objects) conflicted with each other and analysed the flight trajectories of the bees under the different spatial conditions. To test potential effects of optic flow-based guidance systems (such as collision avoidance and flight control) on our conclusion, we presented the bees, in a subsequent experiment, with different wall patterns (low and high contrast as in Chakravarthi et al., 2017) during test conditions.

Experiment set-up

One bumblebee, Bombus terrestris (Linnaeus 1758) hive per experiment, containing a queen, about 30 workers and brood, provided by Koppert B.V. (The Netherlands), was placed in a 35 cm3 plastic box. The bee hive was connected to a tunnel (1.4 m long and with a 30×30 cm2 cross-section) via a 2 cm diameter plastic tube and an 8 cm wide box. The wall and floor of the tunnel were covered with a red and white random pattern following a 1/f1/4 frequency distribution (as in Ravi et al., 2019). The tunnel ceiling consisted of a transparent acrylic panel allowing lighting (350–1000 nm, i.e. part of UV-A to near-infrared) and video recording. A 3 cm long and 1 cm wide platform was placed at the tunnel entrance to facilitate take-off. Bees exiting the nest and entering the tunnel for the first time performed learning flights (Lobecke et al., 2018; Robert et al., 2018). They collided multiple times with the transparent acrylic ceiling of the tunnel until they found the exit at the other end of the tunnel leading to a foraging chamber. Artificial feeders (containing 30% by volume aqueous sugar solution) were placed in the foraging chamber and refilled every second day. The bees had access, via a 20 cm hole in the window of the foraging chamber, to flowers outdoors (Fig. 1). Bees returning to the hive had to cross the same tunnel.

Fig. 1.

Schematic diagram of the experimental set-up from above. Returning bees entered the experimental tunnel from the foraging chamber via a 10 cm diameter tube (experiment 1) or a connecting box (experiment 2). During habituation, four objects were centred on the longitudinal axis of the tunnel (as shown here), and a transparent acrylic plate on either the left or right side of the object forced the bees to fly around the object on the unobstructed side to reach the end of the tunnel and finally enter the hive. Foraging bees were free to forage at artificial feeders (placed in the foraging chamber) or leave the building through a 20 cm diameter hole). The camera (not shown) was placed above the experimental arena. Gates (in blue) were used to regulate the traffic through the tunnel to record one bee at a time and to capture bees for marking (in experiment 2).

Fig. 1.

Schematic diagram of the experimental set-up from above. Returning bees entered the experimental tunnel from the foraging chamber via a 10 cm diameter tube (experiment 1) or a connecting box (experiment 2). During habituation, four objects were centred on the longitudinal axis of the tunnel (as shown here), and a transparent acrylic plate on either the left or right side of the object forced the bees to fly around the object on the unobstructed side to reach the end of the tunnel and finally enter the hive. Foraging bees were free to forage at artificial feeders (placed in the foraging chamber) or leave the building through a 20 cm diameter hole). The camera (not shown) was placed above the experimental arena. Gates (in blue) were used to regulate the traffic through the tunnel to record one bee at a time and to capture bees for marking (in experiment 2).

Natural light came through a large window in the experimental room. Seven slits in the ceiling of the tunnel, aligned with the width of the tunnel and spaced 150 mm apart, allowed the insertion of obstacles. The obstacles inserted through the slits were 2 mm thick plates of transparent acrylic, 20 cm wide and 30 cm high. The obstacles were covered on both sides with a 10 cm coloured paper band (printed on A4 120gsm, ISO536), acting as a visual cue. These stimuli appear orange (hex-code #d49e2a) and purple (hex-code #9e2ad4) to the human eye. The spectral reflectance of the two visual cues was quantified (with a spectrometer OpticOcean4000). The distance between the two colours in the colour space of bumblebees was 0.12 hue (details are given in the accompanying dataset: https://pub.uni-bielefeld.de/record/2945904, see ‘Control’). Hue differences of 0.1 and 0.15 have been shown to be distinguished by bumblebees with approximately 70% and 95% probability, respectively (Chittka, 1992; Dyer and Chittka, 2004). The subtended angle at which the colour cues were seen by a bee 150 mm away from it and 150 mm off the tunnel centre was 19 deg. Bumblebees are able to detect colour contrast for subtended angles larger than 2.7 deg (Dyer et al., 2008). Additionally, the two visual cues, printed on paper, contained achromatic differences. Bumblebees are able to detect long-wavelength receptor contrast for subtended angles larger than 2.3 deg (Dyer et al., 2008). As the cues contained chromatic and achromatic differences, we will refer to them as visual cues (Hempel de Ibarra et al., 2014). Furthermore, the visual cues may not be extracted and processed as single features, but the entire panorama (including the coloured band) may guide the bee along the tunnel (Baddeley et al., 2012; Kodzhabashev and Mangan, 2015; Le Möel and Wystrach, 2020). The obstacles divided the cross-section of the tunnel into three equally sized regions: the coloured band in the centre, flanked on its left and right, respectively, by either an empty region or a transparent acrylic plate. The side of the coloured band on which the acrylic or transparent region was placed depended on the colour covering the central band, therefore acting as a visual direction cue (Fig. 1). Bees could associate the colour (orange and purple) or the panorama before an object with a passing direction (left or right to a colour band, respectively) and/or remember the left/right sequence of manoeuvres to avoid colliding with the transparent acrylic region. Colour cues were the same for both flight directions in the tunnel (returning and exiting bees), i.e. the obstacles had different colours on each side.

During experiments, gates on both sides of the flight tunnel (Fig. 1) were used to regulate traffic and prevent bees from interfering with one another during recordings. Only one bee at a time was permitted to enter the flight tunnel during recordings. Only bees coming back from the foraging chamber in a directed manner, without making loops before entering the tunnel, were recorded. Bees entering in a directed manner are returning experienced foragers, motivated to return to their nest, and all recorded bees returned to the hive.

The bees were given 1 week to get used to travelling through the tunnel without objects before objects were introduced. We observed consistent foraging flights by numerous (>20) worker bees within 1 day of immigration to the enclosure.

The experimental procedure is described in the following from the perspective of the bees returning to the hive.

Procedure

Experiment 1: visual and movement memories

Habituation

Four obstacles were inserted in the slits at −300, −150, 0 and 150 mm relative to the middle of the tunnel. The bees were allowed to travel through the tunnel ad libitum, over three consecutive days. Thus, the traffic of bees was not restricted by gates, and bees may have learnt from one another how to cross the tunnel. The obstacles were placed such that the bees had to alternate between left (indicated by orange) and right (indicated by purple) turns, i.e. had to fly a left–right–left–right sequence. The bees were, thus, punished by collision with the acrylic plate and rewarded by the ability to return home once they managed to pass all four obstacles. After familiarising themselves with the route, bees flew fluently from one end of the tunnel to the other.

Bout procedure

We tested the performance of bumblebees under five different configurations of objects in the tunnel. For each of these test conditions, we recorded foragers returning to their hive in the tunnel. Returning foragers came from the foraging chamber, passed through the tunnel containing the objects, and finally entered the hive. Recordings were done in a sequence of bouts. To ensure that only a single bee was in the tunnel during recordings, we regulated the traffic in the tunnel such that no bees were in the tunnel before the test started. Several bees were trapped in the foraging chamber by closing the gate connecting it to the tunnel. A bout started when the first bee was allowed to enter the tunnel by briefly opening the gate to the feeding chamber. After this bee reached the entrance hole of the hive and was allowed to enter it by opening the corresponding gate, the next bee was allowed to enter from the feeding chamber and to return to the nest entrance at the other end of the tunnel. The bout ended after all bees in the feeding chamber had individually flown back to the hive. Then, both gates were opened, and the traffic in the tunnel was no longer regulated. Between bouts, we placed the objects as in the habituation arrangement (or its mirrored version for the last test condition – see below). We repeated this procedure for each test condition until a sufficient number of flights could be recorded.

Our procedure was thus composed of several bouts of return flights to the nest under one of the test conditions interspersed by unregulated traffic under the habituation condition (or its mirrored version, see above). All flights recorded during a given bout were, of course, performed by different individuals. Up to three bouts were necessary for each test condition to obtain a sufficient number of flights; thus, a given individual may have been recorded maximally 3 times for a given test condition. The strategy used to return home may have differed for the potential bees that were recorded multiple times in a given condition. For example, a bee experiencing the non-alternating sequence (see below) twice may have straightened its path (i.e. followed a path consistent with a visual memory) on a later encounter with this sequence. Such effects were lessened by reintroducing the habitual or mirrored sequence between bouts. The number of tested bees n and the number of bees in the largest bout N, corresponding to the minimal number of individual bees, is given below for the different test conditions.

The date, time and bout of each recording together with one of the following object arrangements in the tunnel are given in the accompanying dataset (https://pub.uni-bielefeld.de/record/2945904, see ‘Summary_exp0.xlsx’).

Four alternating objects

As a reference, we recorded the flights of returning bees in two bouts of 13 and 6 flights, respectively (n=19, N≥13) under the same alternating object configuration that was used for habituation. Thus, the objects between bouts were not changed.

Additional object

The day following the habituation period, a fifth object was introduced at the end of the obstacle sequence to assess whether the bees associate the colour of the object with a turning direction. We recorded the flights of returning bees in three bouts of 4, 6 and 12 flights, respectively (n=22, N≥12).

Acrylic transparency control

Although the acrylic panel was cleaned regularly with ethanol and dried to remove the dust on it, the bees may have been able to detect the acrylic obstacles visually. To test the transparency of the acrylic plate, we presented the bees with an alternating sequence of orange and purple objects, indicating a sequence of left and right turns, respectively. Although the third object indicated a left turn according to its colour, the acrylic panel was placed on the left side. We recorded one bout of 8 bees (Fig. 2E; see also Movie 2), and all collided with the acrylic panel, indicating that the bees were unable to detect the transparent side of the object (n=8, N=8).

Fig. 2.

Trajectories of bees as seen from above for the different object configurations of experiment 1. Trajectories through the tunnel of bees returning from the foraging chamber to the hive are shown in grey. The average trajectory of the bees' individual trajectories is shown in black. The object colour and band location (i.e. the side on which the empty region and transparent acrylic plate are positioned, with the latter indicated by dashed lines) are shown from the perspective of a returning bee. Insets show an example of a bee's position (black line) and body orientation (black ticks) around an object. Distance (mm) is relative to the centre of the tunnel. (A) Four object ‘habituation’ configuration. See Movie 1. (B) An additional fifth object was placed after a habitual sequence of four objects. (C) Mirrored sequence of four objects, i.e. starting with a purple object instead of an orange object. (D) A sequence of six objects without an alternating route-following direction. The number of crosses indicates the number of bees crossing the midline of the tunnel and the object edge (+150 mm for between the two orange objects and −150 mm between the two purple ones), respectively. (E) Control experiment for transparency of the acrylic plate, with the third object indicating a turn toward collision with the transparent acrylic plate. Sections marked in B–D are areas of interest for the quantitative analysis shown in Fig. 3; the bar colours correspond to those in the boxplots in Fig. 3. Note that some trajectories contain gaps due to occlusions from the object barrier or the slit in the tunnel ceiling.

Fig. 2.

Trajectories of bees as seen from above for the different object configurations of experiment 1. Trajectories through the tunnel of bees returning from the foraging chamber to the hive are shown in grey. The average trajectory of the bees' individual trajectories is shown in black. The object colour and band location (i.e. the side on which the empty region and transparent acrylic plate are positioned, with the latter indicated by dashed lines) are shown from the perspective of a returning bee. Insets show an example of a bee's position (black line) and body orientation (black ticks) around an object. Distance (mm) is relative to the centre of the tunnel. (A) Four object ‘habituation’ configuration. See Movie 1. (B) An additional fifth object was placed after a habitual sequence of four objects. (C) Mirrored sequence of four objects, i.e. starting with a purple object instead of an orange object. (D) A sequence of six objects without an alternating route-following direction. The number of crosses indicates the number of bees crossing the midline of the tunnel and the object edge (+150 mm for between the two orange objects and −150 mm between the two purple ones), respectively. (E) Control experiment for transparency of the acrylic plate, with the third object indicating a turn toward collision with the transparent acrylic plate. Sections marked in B–D are areas of interest for the quantitative analysis shown in Fig. 3; the bar colours correspond to those in the boxplots in Fig. 3. Note that some trajectories contain gaps due to occlusions from the object barrier or the slit in the tunnel ceiling.

Fig. 3.

Comparisons of time and straightness of flight between two objects between or within conditions. Boxplots (median, upper and lower quartiles and 95th percentile range; crosses are outliers) compare the two-object sections from Fig. 2B–D. The significance of the difference in mean or median values is indicated: n.s., P>0.05; *P≤0.05; **P≤0.005. Note that it is possible that a given individual was recorded up to 3 times. The assumption of data independence is not fully guaranteed.

Fig. 3.

Comparisons of time and straightness of flight between two objects between or within conditions. Boxplots (median, upper and lower quartiles and 95th percentile range; crosses are outliers) compare the two-object sections from Fig. 2B–D. The significance of the difference in mean or median values is indicated: n.s., P>0.05; *P≤0.05; **P≤0.005. Note that it is possible that a given individual was recorded up to 3 times. The assumption of data independence is not fully guaranteed.

Mirrored sequence

The bumblebees were experienced in passing through a tunnel containing an alternating sequence of objects starting with an orange object indicating a left turn. Thus, the visual memory based on the colour cue and a potential movement memory both indicate the same turning direction. To assess the potential respective contribution of visual and movement memories, we brought them into conflict and evaluated how the bees negotiated the first object after the object sequence was inverted. The first object was now purple instead of orange (Fig. 2C) and indicated a right turn. A bumblebee exclusively following the sequence of movement memories would collide with the acrylic plate, whereas a bumblebee following its visual direction memories would take a collision-free path. We recorded the flights of returning bees in three bouts of 1, 10 and 9 flights, respectively (n=20, N≥10).

After recording the third bout, the mirrored sequence remained in the tunnel and bees were habituated to the mirrored sequence for 2 days before testing the non-alternating sequence to minimise the mixed effect due to some bees having seen the mirror sequence and others not.

Non-alternating sequence

The mirrored sequence only targeted the decision of the bees at the first object. To assess the relevance of the different memories guiding the bees along the route, we placed the objects in a non-alternating (right–left–left–right–right–left) manner (Fig. 2D; see also Movie 3). Similar to the mirrored sequence test, a bumblebee exclusively following the sequence of movement memories would collide with the acrylic plate. In contrast, a bumblebee exclusively following its visual direction memories would keep flying on the left side between two consecutive objects indicating a left turn. Alternatively, a bee guided by combining the two types of memories would have to resolve the conflicting directions given by the different memories and, thus, exhibit more convoluted trajectories and spend more time in front of an object before making a decision. We recorded the flights of returning bees in two bouts of 9 and 11 flights, respectively (n=20, N≥11). Between bouts, the objects were placed as in the mirrored sequence.

Experiment 2: importance of lateral optic flow

Based on the recentring behaviour after an object observed in the first experiment, we concluded that bees were guided by both visual memories and movement memories. Bees control their position and speed in flight tunnels by using estimations of the apparent motion of objects (i.e. optic flow) (Srinivasan et al., 1996; Linander et al., 2015; Portelli et al., 2017; Lecoeur et al., 2018). These estimations are probably acquired by elementary motion detectors, their response being sensitive to the contrast of the objects' texture (Egelhaal et al., 2014). Chakravarthi et al. (2017) have shown that the texture on a tunnel's walls affects the bumblebee’s flying position in the tunnel, with a low-contrast and high-contrast pattern yielding low and high estimations, respectively. Thus, to investigate the interaction between optic flow, visual memories and movement memories, we conducted a post hoc experiment.

Habituation

The bees were allowed to travel ad libitum through an empty tunnel. Bees returning swiftly to the tunnel were captured in the connector between the hive and the tunnel (Fig. 1) and individually marked with a numbered and coloured marker waxed on their thorax while restraining the bee. The bees were allowed to travel through the tunnel ad libitum over three consecutive days, as in experiment 1.

Testing

We challenged the bees to cross a tunnel with low- or high-contrast texture on the wall and a non-alternating visual object sequence. The low-contrast texture was a white pattern. The high-contrast texture was a black/white sinewave grating (at 0.5 cycles cm−1). If the recentring behaviour were mainly due to the response to the apparent motion of the wall on the retina, the bee would recentre more at high contrast than at low contrast. Twelve returning marked bees, identified within the connector between the foraging chamber and the tunnel, were tested once in each sequence. Bees were recorded one at a time, and conditions were chosen based on the bee ID. Once a bee could be tested under both conditions, it was removed from the hive.

Flight trajectory analyses

A camera (Optronis CR3000×2 for experiment 1, Basler acA 2040 µm-NIR for experiment 2) was placed 1.5 m above the midline of the flight tunnel pointing directly downwards. The flights of the bees were recorded at 60 Hz under the different test conditions as explained above.

Trajectories from camera

We estimated the camera position and orientation relative to the tunnel and the camera lens parameters by using a direct linear transformation (Hedrick, 2008) on points with known positions within the tunnel. We wrote a custom-made Python (http://www.python.org) code to process the video footage, allowing the fit of an ellipse around the bee to compute its location and long-axis orientation. Bees tend to fly close to the entrance and exit altitude of the tunnel (Portelli et al., 2010). Therefore, we assumed a constant height of 150 mm as a proxy for our data analysis. Using a different height would have slightly changed the reconstructed positions of the bees within the tunnel but would not have affected our overall conclusions (see https://pub.uni-bielefeld.de/record/2945904, ‘Control’).

Flight characteristics

The flight trajectories and orientation of the bees were 8th order low-pass filtered with a 10 Hz cut-off frequency to attenuate digitizing errors due to a slight variation in determining the centre of mass of the bee and the orientation of the ellipse. Heading orientation was described by the angle between the bee long axis and the tunnel long axis. A bee facing the tunnel exit (toward the hive) had a null orientation. A bee facing the left wall had a positive 90 deg orientation, and a bee facing the right wall had a negative 90 deg orientation. Our convention thus follows the right-hand rule. The flight velocity was determined within a bee-centred coordinate system. The flight time and straightness of the flight trajectories were calculated between two consecutive relevant obstacles (i.e. a relevant tunnel section). The inverse of the sinuosity of flight trajectories (later called straightness) was obtained by dividing the shortest distance (the Euclidean distance between the start and end of the trajectory) by the distance travelled (estimated by integrating the speed along the trajectory) (Benhamou, 2004).

Statistical comparison

We quantified the flights of the bees between two consecutive relevant obstacles (i.e. tunnel sections) by calculating the flight time and the straightness between these two obstacles. The flight characteristics were compared between key sections across and within object configurations. The bees were unmarked in the first experiment, and we used unpaired statistical tests to the compare flight characteristics between sections in different object configurations. Paired tests were used to compare differences in the same object configuration. The flight times and the straightness of flight trajectories are bounded measures ([0,∞] and [0,1], respectively) and, thus, the data may not be normally distributed. Therefore, we assessed the departure from normality by using the D'Agnostino–Pearson test. We considered that the data departed from a normal distribution for P-values smaller than 0.05. Thus, non-parametric tests were used when the departure from normality was significant. We compared the distributions of the straightness between sections (and times within a section) within an object configuration and in different object configurations by using a Wilcoxon test (Fig. 3A) and Mann–Whitney U-test (Fig. 3B,C), respectively. While the P-value provided by the statistical test can inform whether an effect exists, it does not reveal the size of the effect (Sullivan and Feinn, 2012). One method of reporting the effect size for the Wilcoxon test and Mann–Whitney U-test is the common language effect size. This was computed by forming all possible pairs between the two groups and then finding the proportion of pairs that supports the hypothesis that bumblebees flew along a straighter path in one object configuration compared to the other (McGraw and Wong, 1992).

Regarding the non-alternating direction section (Fig. 2D), the bees first recentred and then followed a collision-free path. In order to compare the recentring direction with the direction given by a path integrator, we determined the bee's flight vector at the entrance of a non-alternating section (i.e. at 150 and −150 mm for the first and second non-alternating locations, respectively) by calculating the difference between their positions in the tunnel after travelling 60 mm past the entrance of a non-alternating section and their positions at the entrance of the section. The path integrator vector was the difference between the nest entrance and the bee's position at the entrance of the section. The vectors were expressed in polar coordinates in order to extract the flight direction of the bee and the direction of the path integrator. The two directions did not appear to be skewed and were unimodal. Therefore, we compared the two directions with the parametric Hotelling paired sample test (see section 27.14 of Zar, 2006). The recentring direction may be compared with the habitual direction in the condition ‘habituation’ during left and right turns. After extracting the habitual direction (similar to the recentring direction in the non-alternating location), we compared the habitual direction and the recentring direction by using a parametric Watson–Williams multi-sample test. Although the Watson–Williams tests assume that each of the samples comes from a von Mises distribution (circular analog to the normal distribution of linear data), it is robust to departures from these assumptions as long as the assumptions are not severely violated (e.g. when the distributions are not unimodal) (see section 27.5 of Zar, 2006).

Ten statistical tests were performed in the first experiment. The greater the number of statistical tests, the more likely it is to find a given test that yields significant results (a problem known as the multiple comparisons problem). We used a stricter significance threshold for individual comparisons than P=0.05 to prevent this from happening. The Bonferroni correction suggests using a threshold inversely proportional to the number of tests performed. Therefore, we used a statistical threshold of 0.005.

In the second experiment, we quantified the recentring by measuring the maximum deviation from the wall in the non-alternating sections. The deviations were expressed relative to the centre of the tunnel. Because the number of samples was smaller than 20, we assessed departures from normality by using Shapiro–Wilk test. We compared the marked bees in two conditions with a paired t-test and Wilcoxon test for parametric and non-parametric distribution, respectively. The Cohen and common language effect size were also calculated. Four statistical tests were performed in this experiment. Following Bonferroni correction, we used a statistical threshold of 0.0125.

Average trajectory

The average trajectory was calculated as follows to visualize the average bee behaviour for the different object configurations (e.g. black line in Fig. 2A). As the bees did not fly at exactly the same speed, the trajectories needed to be temporarily aligned. One trajectory in the set of trajectories for a given condition was selected randomly and used as a reference for temporal alignment. All remaining trajectories within this condition were then temporally aligned by using dynamic time warping (an algorithm used to measure similarity between two temporal sequences varying in speed; Salvador and Chan, 2007). The time point along the trajectory corresponded to at least one time point in the reference trajectory for any given trajectory, i.e. ‘matched’ a point in the reference trajectory (multiple time points can correspond to the same time point in the reference trajectory). The average position of all matched points was calculated along the time of the reference trajectory.

All analyses were done in Python 3.7. A list of the packages used is given in the accompanying dataset (https://pub.uni-bielefeld.de/record/2945904).

Combined motor and visual learning

Which navigational guidance systems might be at work in steering bumblebees toward their home? Bumblebees were habituated to return home along a route formed by four objects, with these visual cues indicating alternately left and right turns. After habituation, the bees could potentially rely on visual or movement memories or a combination of the two. An additional object was added at the end of their habitual route to assess whether the bees generalise their strategy to novel situations. The bees followed the direction indicated by the visual cues carried by the object. Although we did not find statistical differences in the time or the straightness between a habitual route section and the additional section (Fig. 3A), in three flights, collisions (one in the first bout, and two in the third bout) were observed with the centre of the objects (see individual trajectories in Fig. 2B), indicating that at least some bees were not expecting an additional object and, thus, might have followed their movement memories (i.e. flying in the middle of the tunnel).

To investigate whether bumblebees used movement memories in addition to visual memories, we challenged them with a novel object configuration that was mirrored compared with their habitual route. The first object along this novel path thus indicated a right turn instead of the habitual left turn (Fig. 2C). The flight trajectories were more spread out, and bees spent more time in front of the first object in the mirrored condition compared with the habituated one (Mann–Whitney U-test: P<0.002 and P<0.0002, effect size: 0.77 and 0.17, respectively; Fig. 3B). One bee even turned left and, thus, collided with the transparent acrylic plate. This behaviour might have been caused by the bumblebees getting used to a left turn occurring first (movement memories) in the ‘habituation’ configuration. However, the bees might have also built up an expectation regarding which colour cue appears first. Regardless, they did not follow the direction indicated by the object colour exclusively and, thus, the visual rule provided by the object colour.

The flight tunnel was modified in another way to investigate further whether movement memories play a role in addition to visual memories when navigating back to the nest. With a non-alternating sequence of objects (Fig. 2D), most bees recentred in the tunnel as if they were used to alternating between turns, as in the ‘habituation’ tunnel configuration. This behaviour is expected if they were using movement memories in addition to the visual cues. We quantified the proportion of flights with recentring behaviour at the conflict location (indicating the presence of movement memories), by classifying the trajectories of the bees into two categories: (1) following the wall, i.e. no recentring (left side, y>50) and (2) crossing the object edge, i.e. recentring (location from tunnel centre: y<50). We found that only a quarter of the trajectories recorded stayed on the left of the tunnel. Those bees appeared to be driven primarily by visual cues and decided early on to stay on the left side of the tunnel. However, the rest of the bees recorded (i.e. three-quarters of them) flew at least for some time along a path that could be explained by following movement memories, i.e. by a habitual changing of tunnel side after an object. These bees may alternatively need to have the visual cues at the centre of their visual field in order to elicit the visually triggered turn.

We determined the straightness of the trajectories and the time spent between two identical objects and compared these characteristics with the trajectories between two consecutive objects along the habitual route to quantify how flight trajectories were affected by the non-alternating object sequence. We did not find statistical differences in either straightness or time (Mann–Whitney U-test, P>0.12, effect size: 0.85; Fig. 3C). Thus, despite the conflict situation, the bee resolves the conflict quickly and chooses to follow the visual direction.

Potential role of path integration

Navigating insects may also use path integration to return home. The path integrator yields a vector pointing from the current position of the bee to the start of its journey (here, the nest entrance at one end of the flight tunnel). When the bee is on the left side (or right side) of the tunnel, the path integrator would indicate a direction slightly to the right (or correspondingly left). Thus, the recentring behaviour observed above could potentially also result from a path integrator.

We compared the direction the bees flew 60 mm past the second (or correspondingly fourth) object along the non-alternating route, with the direction given by the path integrator (Fig. 4) to assess whether bumblebees' recentring behaviour was due to the path integrator. The two directions differed significantly from each other at the first occurrence of the non-alternating objects, but not at the second (parametric Hotelling paired sample test: P<0.0001 and P=0.01 at the first and second non-alternating directions, respectively). We compared the bees' directions with those taken during the first object configuration (i.e. after the second and third object of Fig. 2A) and found that they did not differ statistically (parametric Watson–Williams multi-sample test: P=0.43 and P=0.27 at the first and second non-alternating directions, respectively). The path integrator was, therefore, concluded to be a less plausible explanation of the recentring behaviour than the replication of habitual movement.

Fig. 4.

Comparison of recentring behaviour with path integration and habitual movement directions. Trajectories are shown in blue relative to the first and second non-alternating location of two objects (i.e. A at 150 mm, B at −150 mm) in the non-alternating object configuration (see Fig. 2D). The direction of path integration at the beginning of the trajectories is in orange. The average trajectories (from the habituation sequence) for a right (A) and left turn (B) are plotted in black. The dots indicate flight direction in the same colour as the lines. The horizontal bar and curved line indicate the mean and standard deviation of the distribution of directions, respectively (see Materials and Methods for statistical comparison). Note that it is likely that a given individual was recorded multiple times.

Fig. 4.

Comparison of recentring behaviour with path integration and habitual movement directions. Trajectories are shown in blue relative to the first and second non-alternating location of two objects (i.e. A at 150 mm, B at −150 mm) in the non-alternating object configuration (see Fig. 2D). The direction of path integration at the beginning of the trajectories is in orange. The average trajectories (from the habituation sequence) for a right (A) and left turn (B) are plotted in black. The dots indicate flight direction in the same colour as the lines. The horizontal bar and curved line indicate the mean and standard deviation of the distribution of directions, respectively (see Materials and Methods for statistical comparison). Note that it is likely that a given individual was recorded multiple times.

Potential role of lateral optic flow

Bumblebees and other flying insects rely heavily on optic flow to avoid collision (Kern et al., 2012; Lecoeur et al., 2019), maintain a direction (Portelli et al., 2010), control their speed (Baird et al., 2005), negotiate gaps (Ravi et al., 2019) and prepare for landing (Frasnelli et al., 2018). They tend to fly close to the centre in tunnels (Srinivasan et al., 1996; Linander et al., 2015) or along the shortest route to food sources by balancing the optic flow perceived in their left and right visual fields (Portelli et al., 2017). The recentring behaviour described above could be due to balancing the optic flow in their two eyes.

To challenge this possibility, we habituated bumblebees with a random pattern on the lateral walls of the flight tunnel with the ‘habituation’ object configuration (as in Fig. 2A). We then tested individuals with non-alternating object sequences (as in Fig. 2D). However, in contrast to the experiments performed under the non-alternating object conditions, the contrast of the wall pattern was either high or low (see Fig. 5A,B). We classified the trajectories of the bees into two categories, as was done for the data analysis shown in Fig. 2D, i.e. staying on one side or crossing the object's edge. We found that 3 out of 12 of the recorded trajectories with a high-contrast wall pattern (and 5 out 12 with the low-contrast wall pattern) stayed on the right side of the tunnel. However, the rest of the bees recorded flew, at least for some time, along a path predicted if they were using movement memories, i.e. by changing tunnel side after an object.

Fig. 5.

Trajectories of bees seen from above in a flight tunnel with non-alternating sequences with varying optic flow on the wall. The wall pattern provided high optic flow estimates (A) and low optic flow estimates (B) to bees returning from the foraging chamber to the hive; bees were previously habituated with the four object configuration (Fig. 2A). Box plots (median, upper and lower quartiles and 95th percentile range; crosses are outliers) show the recentring strength of the bees at the first (Ci) and second (Cii) non-alternating location of two objects (i.e. [150,0 mm] and [−150,−300 mm], respectively) in the two conditions (high and low optic flow). n.s., P>0.0125; **P≤0.0025.

Fig. 5.

Trajectories of bees seen from above in a flight tunnel with non-alternating sequences with varying optic flow on the wall. The wall pattern provided high optic flow estimates (A) and low optic flow estimates (B) to bees returning from the foraging chamber to the hive; bees were previously habituated with the four object configuration (Fig. 2A). Box plots (median, upper and lower quartiles and 95th percentile range; crosses are outliers) show the recentring strength of the bees at the first (Ci) and second (Cii) non-alternating location of two objects (i.e. [150,0 mm] and [−150,−300 mm], respectively) in the two conditions (high and low optic flow). n.s., P>0.0125; **P≤0.0025.

We measured how far the bees were moving relative to the tunnel midline to quantify the strength of recentring. When comparing the recentring strength for the high- and low-contrast tunnel walls at the first non-alternating location along the route, we did not find significant differences (Wilcoxon: P=0.14). However, the same comparison yielded significant results (Wilcoxon: P<0.004) at the second non-alternating location (Fig. 5C). Therefore, the optic flow estimates do not seem to be the main factor pushing bees toward the centre of the tunnel. However, they play a role after a conflict between movement and visual memories has been encountered.

Control of movement memories

To investigate whether optic flow played a role in shaping the manoeuvres around the obstacles, we compared the direction the bees flew 60 mm past the second object along the non-alternating route in the condition with high- and low-contrast wall patterns, yielding high and low optic flow estimations, respectively (Fig. 6). The two directions differed significantly from each other at the first occurrence of non-alternating objects (parametric Hotelling paired sample test: P<0.001). Thus, optic flow may be used to shape the fine structure of habitual manoeuvres.

Fig. 6.

Comparison of recentring behaviour with two different strengths of optic flow. The blue and orange lines respectively depict the trajectories relative to the first and second non-alternating location of two objects (A at 150 mm, B at −150 mm) in the non-alternating object configuration (see Fig. 5) in the high (A) and low optic flow (B) conditions. The dots indicate flight direction in the same colour as the lines. The horizontal bar and curved line indicate the mean and standard deviation of the distribution of directions, respectively (see Materials and Methods for statistical comparison).

Fig. 6.

Comparison of recentring behaviour with two different strengths of optic flow. The blue and orange lines respectively depict the trajectories relative to the first and second non-alternating location of two objects (A at 150 mm, B at −150 mm) in the non-alternating object configuration (see Fig. 5) in the high (A) and low optic flow (B) conditions. The dots indicate flight direction in the same colour as the lines. The horizontal bar and curved line indicate the mean and standard deviation of the distribution of directions, respectively (see Materials and Methods for statistical comparison).

Integration of movement memories and visual memories

While navigating along previously learnt habitual routes, ants and bees are guided by multiple systems relying on different types of memories, be they visual (Zeil, 2012; el Jundi et al., 2016; Wystrach et al., 2011; Menzel et al., 2019), olfactory (Buehlmann et al., 2014), path integration (Pfeffer and Wittlinger, 2016; Wittlinger et al., 2006) or sequence of movement memories (Zhang et al., 1996, 2012; Macquart et al., 2008; Mirwan and Kevan, 2015). In the present study, the bees flew along a habitual route alternating left and right turns. Visual guidance (orange and purple objects), movement guidance (the sequence of turns) or a mixture of the two could steer the bees along this route. When the visual cues were placed to create a conflict between the visual and movement guidance, the bees flew on the side of the object indicated by the visual cue. However, we observed that prior to this decision, bumblebees recentred to the middle of the tunnel. This recentring may be the signature of a movement guidance system or the need to recentre to process the visual cues although the cues were large enough to be processed at the location of the previous object. By inverting the sequence, we observed that bees tended to fly along their habitual path (i.e. left turn). In this situation, the cue was in front of the bee and therefore the recentring, observed with the non-alternating sequence, may be the signature of a movement guidance system, possibly tuned by an interplay of lateral optic flow and path integration. The bees, thus, would follow the route by a weighted combination of movement and visual guidance system. But could these be a single guidance system?

It may well be that the bees do not associate the colour with the choice regarding which side of the object to fly around, but associate the visual information with a complete chain of manoeuvres (or motor patterns). This chain of manoeuvres, for example, could be the complete path around a single object ending at the tunnel midline, where the next chain of manoeuvres can then start (e.g. a left or a right path). In our experiment, the bees would pass a purple object on the left, finishing the learnt manoeuvre in front of the next object. At this point, the bee would start the second manoeuvre, indicated by the object colour lying in the bee's path. In the case of conflict, the object would be purple again; thus, it would initiate the same chain of manoeuvres. Hence, our observed recentring in the conflict situation could also result from this visually triggered chain of manoeuvres around objects. Thus, the movement and visual guidance will not be continuously integrated, but visual guidance will dictate which movement to perform.

Are bees guided by integration of movement and visual guidance or by a visually triggered chain of manoeuvres? Our experimental design does not allow us to answer this question directly, but several researchers have observed movement memories, and we will discuss the implications for our two guidance suggestions: (1) visual and movement memories that are continuously integrated (i.e. integrated guidance system), and (2) a chain of manoeuvres that is visually triggered (i.e. triggered guidance system).

The visual cues in our indoor experimental paradigm were clearly distinguishable, easing the association with a movement memory. However, visual cues may not be as evident in natural environments. Thus, in the absence of other guidance systems, the insects may be guided by movement memories. Indeed, bees and ants are able to replicate a whole chain of manoeuvres in visually bare mazes (Collett et al., 1993; Zhang et al., 1996, 2012; Macquart et al., 2008; Mirwan and Kevan, 2015). The visual information is not present within the maze but only at its entrance. Within the maze, a turn made by an individual is based on the sequence of previous turns (for example, after turning left, turn right) and not on the current visual information. An individual may recognise the sequence of manoeuvres to perform by using visual features at the entrance of the maze (visually triggered guidance system). In contrast, the individual may reach the maze entrance by using an integration of visual and movement memories, and continue to use such a strategy despite the absence of visual features within the maze. In the absence of visual memories, an integrated guidance system becomes equivalent to a system based solely on movement memories. In other words, an individual replicating the entire movement sequence (triggered guidance system) is equivalent to a bee solely guided by movement memories (integrated guidance system). These findings suggest that bumblebees in our set-up may have learnt the complete sequence of turns in the tunnel.

Collett et al. (1993) found that when visual cues are clearly present, the sequence of movement is not replicated. In their experiment, the bees had to enter a compartment with a visual cue placed on the rear of the entrance and then decide which of two exits to take. After entering the compartment, the bees turned around to view the visual cue and then flew through the exit indicated; thus, they did not replicate the movement sequence. In our integrated guidance suggestion, this would mean that the movement guidance system is strongly down-weighted to a point where it is no longer observable. Collett et al.’s (1993) finding could also be explained by a triggered guidance system, where the visual cue on the back of the entrance triggered a chain of manoeuvres guiding the bee through the left or right exit.

The two guidance suggestions are, thus, plausible explanations for the recentring behaviour that we observed. Nevertheless, they rely on different mechanisms. The integration of several guidance systems in insect navigation has been modelled based on the certainty of cues (e.g. path integration and collision avoidance: Bertrand et al., 2015; and path integration and visual guidance: Hoinville and Wehner, 2018). In this framework, each guidance system provides a direction and a certainty to follow this direction. The directions are then integrated based on the certainties of the different guidance systems. Incorporating movement memories into this framework would require that the movement guidance yields not only a direction but also a certainty, i.e. requires a certainty function. Therefore, it would be interesting to investigate such a function, as it may change with the complexity of the chain of manoeuvres. The trigger of a chain of actions has been observed multiple times, be it fixed action patterns (e.g. Lorenz, 1981) or learned action patterns (e.g. Matsumoto et al., 2012; Lichtenstein et al., 2018). Akin to the proboscis extension reflex, it would be interesting to investigate whether other action patterns (such as a sequence of manoeuvres) that can play a role during navigation (e.g. flying to the left) can be triggered by visual cues.

While previously movement memories were not often considered as an explanation for the path taken by bees during a navigational task, we now provide evidence that movement and visual guidance may interact when following a route and, therefore, raise the question of whether this guidance system plays a role in other navigational contexts, such as reaching a feeder, pinpointing the nest entrance, or when facing a novel situation, for example after a change in the visual surroundings during experiments.

Potential importance of experience in tuning guidance systems

Along a regular route, travelled multiple times, both the scenery between the subsequent journeys and the manoeuvres to be performed would be familiar. With our experiments, we wanted to investigate whether bumblebees remembered more than the visual scenery along a habitual route.

Indeed we found that in some cases, not only were visual cues paving the route followed but also a behaviour consistent with a guidance system based on movement memories was observed.

In the first experiment, we found that 15 flights out of the 20 recorded were following habitual movements before the bees adjusted their path according to the visual cue. However, we do not know the prior experience of the recorded bees. It may be well the case that some bees adjust the balance between the two guidance strategies with experience, i.e. do not recentre in the non-alternating condition when experiencing it for the second time. Some of the bees recorded may have experienced a non-alternating condition (e.g. the acrylic control) before flying through the non-alternating test. We thus cannot unambiguously assess whether the balance between the two guidance systems changes with experience from the first experiment.

However, in the second experiment, the bees were individually marked. They were tested twice in a non-alternating condition, one with a low-contrast pattern on the wall, the other with a high-contrast pattern. We found that recentring was observed in both conditions. So, do the bees behave in a different manner in their first encounter with a non-alternating condition from that in their second? Of the 10 bees that recentred in at least one condition, we found six bees that recentred in both conditions, two bees that only recentred in the first encounter with the non-alternating condition, and two bees that only recentred in the second encounter. Some bees, thus, seem to adjust the balance between the two guidance strategies.

As we did not record the entire history or experience of the bees with the habitual route, we cannot relate the different observed behaviours with the bees’ experience or their individuality. How the balance between the two guidance systems emerges as a result of training, experience or individual preference is beyond the scope of our study and thus remains to be answered.

We would like to thank Luise Odenthal for her technical support during the analyses, Andrea Gonsek, Magdalena Rados and Silvia Rönnau for their help during experiments, and the reviewers for their helpful criticisms.

Author contributions

Conceptualization: O.J.N.B., C.D., T.S., S.R.; Methodology: O.J.N.B., T.S., S.R.; Software: O.J.N.B.; Validation: C.D., M.E.; Formal analysis: O.J.N.B., C.D.; Investigation: O.J.N.B., T.S.; Resources: O.J.N.B., M.E.; Data curation: O.J.N.B.; Writing - original draft: O.J.N.B.; Writing - review & editing: O.J.N.B., C.D., S.R., M.E.; Visualization: O.J.N.B.; Supervision: O.J.N.B., S.R.; Project administration: M.E.; Funding acquisition: M.E.

Funding

The Deutsche Forschungsgemeinschaft (DFG) supported the project financially and a Research Fellowship was awarded to S.R. from the Alexander von Humboldt Stiftung.

Data availability

The data and code that support the findings of this study are openly available from Bielefeld University: https://pub.uni-bielefeld.de/record/2945904.

Baddeley
,
B.
,
Graham
,
P.
,
Husbands
,
P.
and
Philippides
,
A.
(
2012
).
A model of ant route navigation driven by scene familiarity
.
PLoS Comput. Biol.
8
,
e1002336
.
Baird
,
E.
,
Srinivasan
,
M. V.
,
Zhang
,
S.
and
Cowling
,
A.
(
2005
).
Visual control of flight speed in honeybees
.
J. Exp. Biol.
208
,
3895
-
3905
.
Benhamou
,
S.
(
2004
).
How to reliably estimate the tortuosity of an animal's path: straightness, sinuosity, or fractal dimension?
J. Theor. Biol.
229
,
209
-
220
.
Bertrand
,
O. J. N.
,
Lindemann
,
J. P.
and
Egelhaaf
,
M.
(
2015
).
A bio-inspired collision avoidance model based on spatial information derived from motion detectors leads to common routes
.
PLoS Comput. Biol.
11
,
e1004339
.
Bregy
,
P.
,
Sommer
,
S.
and
Wehner
,
R.
(
2008
).
Nest-mark orientation versus vector navigation in desert ants
.
J. Exp. Biol.
211
,
1868
-
1873
.
Buatois
,
A.
and
Lihoreau
,
M.
(
2016
).
Evidence of trapline foraging in honeybees
.
J. Exp. Biol.
219
,
2426
-
2429
.
Buehlmann
,
C.
,
Graham
,
P.
,
Hansson
,
B. S.
and
Knaden
,
M.
(
2014
).
Desert ants locate food by combining high sensitivity to food odors with extensive crosswind runs
.
Curr. Biol.
24
,
960
-
964
.
Chakravarthi
,
A.
,
Kelber
,
A.
,
Baird
,
E.
and
Dacke
,
M.
(
2017
).
High contrast sensitivity for visually guided flight control in bumblebees
.
J. Comp. Physiol. A
203
,
999
-
1006
.
Chittka
,
L.
(
1992
).
The colour hexagon: a chromaticity diagram based on photoreceptor excitations as a generalized representation of colour opponency
.
J. Comp. Physiol. A
170
,
533
-
543
.
Collett
,
M.
(
2012
).
How navigational guidance systems are combined in a desert ant
.
Curr. Biol.
22
,
927
-
932
.
Collett
,
T.
,
Fry
,
S.
and
Wehner
,
R.
(
1993
).
Sequence learning by honeybees
.
J. Comp. Physiol. A
172
,
693
-
706
.
Dyer
,
A. G.
and
Chittka
,
L.
(
2004
).
Biological significance of distinguishing between similar colours in spectrally variable illumination: bumblebees (Bombus terrestris) as a case study
.
J. Comp. Physiol. A Neuroethol. Sensory Neural. Behav. Physiol.
190
,
105
-
114
.
Dyer
,
A. G.
,
Spaethe
,
J.
and
Prack
,
S.
(
2008
).
Comparative psychophysics of bumblebee and honeybee colour discrimination and object detection
.
J. Comp. Physiol. A Neuroethol. Sensory. Neural Behav. Physiol.
194
,
617
-
627
.
Egelhaal
,
M.
,
Kern
,
R.
and
Lindemann
,
J. P.
(
2014
).
Motion as a source of environmental information: a fresh view on biological motion computation by insect brains
.
Front. Neural Circuits
8
,
127
.
el Jundi
,
B.
,
Foster
,
J. J.
,
Khaldy
,
L.
,
Byrne
,
M. J.
,
Dacke
,
M.
and
Baird
,
E.
(
2016
).
A snapshot-based mechanism for celestial orientation
.
Curr. Biol.
26
,
1456
-
1462
.
Frasnelli
,
E.
,
Hempel de Ibarra
,
N.
and
Stewart
,
F. J.
(
2018
).
The dominant role of visual motion cues in bumblebee flight control revealed through virtual reality
.
Front. Physiol.
9
,
1038
.
Hedrick
,
T. L.
(
2008
).
Software techniques for two- and three-dimensional kinematic measurements of biological and biomimetic systems
.
Bioinspir. Biomim.
3
,
034001
.
Heinze
,
S.
,
Narendra
,
A.
and
Cheung
,
A.
(
2018
).
Principles of insect path integration
.
Curr. Biol.
28
,
R1043
-
R1058
.
Hempel de Ibarra
,
N.
,
Vorobyev
,
M.
and
Menzel
,
R.
(
2014
).
Mechanisms, functions and ecology of colour vision in the honeybe
.
J. Comp. Physiol. A Neuroethol. Sens. Neural Behav. Physiol.
200
,
411
-
433
.
Hoinville
,
T.
and
Wehner
,
R.
(
2018
).
Optimal multiguidance integration in insect navigation
.
Proc. Natl Acad. Sci. USA
115
,
2824
-
2829
.
Kern
,
R.
,
Boeddeker
,
N.
,
Dittmar
,
L.
and
Egelhaaf
,
M.
(
2012
).
Blowfly flight characteristics are shaped by environmental features and controlled by optic flow information
.
J. Exp. Biol.
215
,
2501
-
2514
.
Kodzhabashev
,
A.
and
Mangan
,
M.
(
2015
).
Route following without scanning
. In
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
, vol.
9222
, pp.
199
-
210
.
Springer Verlag
.
Kohler
,
M.
and
Wehner
,
R.
(
2005
).
Idiosyncratic route-based memories in desert ants, Melophorus bagoti: how do they interact with path-integration vectors?
Neurobiol. Learn. Mem.
83
,
1
-
12
.
Le Möel
,
F.
and
Wystrach
,
A.
(
2020
).
Opponent processes in visual memories: a model of attraction and repulsion in navigating insects’ mushroom bodies
.
PLoS Comput. Biol.
16
,
e1007631
.
Lecoeur
,
J.
,
Baird
,
E.
and
Floreano
,
D.
(
2018
).
Spatial encoding of translational optic flow in planar scenes by elementary motion detector arrays
.
Sci. Rep.
8
,
5821
.
Lecoeur
,
J.
,
Dacke
,
M.
,
Floreano
,
D.
and
Baird
,
E.
(
2019
).
The role of optic flow pooling in insect flight control in cluttered environments
.
Sci. Rep.
9
,
7707
.
Legge
,
E. L. G.
,
Wystrach
,
A.
,
Spetch
,
M. L.
and
Cheng
,
K.
(
2014
).
Combining sky and earth: desert ants (Melophorus bagoti) show weighted integration of celestial and terrestrial cues
.
J. Exp. Biol.
217
,
4159
-
4166
.
Lichtenstein
,
L.
,
Lichtenstein
,
M.
and
Spaethe
,
J.
(
2018
).
Length of stimulus presentation and visual angle are critical for efficient visual PER conditioning in the restrained honey bee, Apis mellifera
.
J. Exp. Biol.
221
,
jeb179622
.
Linander
,
N.
,
Baird
,
E.
and
Dacke
,
M.
(
2015
).
Bumblebee flight performance in environments of different proximity
.
J. Comp. Physiol. A Neuroethol. Sensory Neural Behav. Physiol.
202
,
97
-
103
.
Lipp
,
H.-P.
,
Vyssotski
,
A. L.
,
Wolfer
,
D. P.
,
Renaudineau
,
S.
,
Savini
,
M.
,
Tröster
,
G.
and
Dell'Omo
,
G.
(
2004
).
Pigeon homing along highways and exits
.
Curr. Biol.
14
,
1239
-
1249
.
Lobecke
,
A.
,
Kern
,
R.
and
Egelhaaf
,
M.
(
2018
).
Taking a goal-centred dynamic snapshot as a possibility for local homing in initially naïve bumblebees
.
J. Exp. Biol.
221
,
jeb168674
.
Lorenz
,
K. Z.
(
1981
).
The centrally coordinated movement or fixed motor pattern
. In
The Foundations of Ethology
, pp.
107
-
152
.
Vienna
:
Springer Vienna
.
Macquart
,
D.
,
Latil
,
G.
and
Beugnon
,
G.
(
2008
).
Sensorimotor sequence learning in the ant Gigantiops destructor
.
Anim. Behav.
75
,
1693
-
1701
.
Matsumoto
,
Y.
,
Menzel
,
R.
,
Sandoz
,
J.-C.
and
Giurfa
,
M.
(
2012
).
Revisiting olfactory classical conditioning of the proboscis extension response in honey bees: a step toward standardized procedures
.
J. Neurosci. Methods
211
,
159
-
167
.
McGraw
,
K. O.
and
Wong
,
S. P.
(
1992
).
A common language effect size statistic
.
Psychol. Bull.
111
,
361
-
365
.
Menzel
,
R.
,
Tison
,
L.
,
Fischer-Nakai
,
J.
,
Cheeseman
,
J.
,
Balbuena
,
M. S.
,
Chen
,
X.
,
Landgraf
,
T.
,
Petrasch
,
J.
,
Polster
,
J.
and
Greggers
,
U.
(
2019
).
Guidance of navigating honeybees by learned elongated ground structures
.
Front. Behav. Neurosci.
12
,
1
-
14
.
Mirwan
,
H. B.
and
Kevan
,
P. G.
(
2015
).
Maze navigation and route memorization by worker bumblebees (Bombus impatiens (Cresson) (Hymenoptera: Apidae)
.
J. Insect Behav.
28
,
345
-
357
.
Pfeffer
,
S. E.
and
Wittlinger
,
M.
(
2016
).
Optic flow odometry operates independently of stride integration in carried ants
.
Science
353
,
1155
-
1157
.
Portelli
,
G.
,
Ruffier
,
F.
and
Franceschini
,
N.
(
2010
).
Honeybees change their height to restore their optic flow
.
J. Comp. Physiol. A
196
,
307
-
313
.
Portelli
,
G.
,
Serres
,
J. R.
and
Ruffier
,
F.
(
2017
).
Altitude control in honeybees: joint vision-based learning and guidance
.
Sci. Rep.
7
,
9231
.
Ravi
,
S.
,
Bertrand
,
O.
,
Siesenop
,
T.
,
Manz
,
L.-S.
,
Doussot
,
C.
,
Fisher
,
A.
and
Egelhaaf
,
M.
(
2019
).
Gap perception in bumblebees
.
J. Exp. Biol.
222
,
jeb184135
.
Robert
,
T.
,
Frasnelli
,
E.
,
Hempel De Ibarra
,
N.
and
Collett
,
T. S.
(
2018
).
Variations on a theme: bumblebee learning flights from the nest and from flowers
.
221
,
jeb172601
.
Salvador
,
S.
and
Chan
,
P.
(
2007
).
Toward accurate dynamic time warping in linear time and space
.
Intell. Data Anal.
11
,
561
-
580
.
Srinivasan
,
M.
,
Zhang
,
S.
,
Lehrer
,
M.
and
Collett
,
T.
(
1996
).
Honeybee navigation en route to the goal: visual flight control and odometry
.
J. Exp. Biol.
199
,
237
-
244
.
Sullivan
,
G. M.
and
Feinn
,
R.
(
2012
).
Using effect size-or why the P value is not enough
.
J. Grad. Med. Educ.
4
,
279
-
282
.
Wittlinger
,
M.
,
Wehner
,
R.
and
Wolf
,
H.
(
2006
).
The ant odometer: stepping on stilts and stumps
.
Science
312
,
1965
-
1967
.
Woodgate
,
J. L.
,
Makinson
,
J. C.
,
Lim
,
K. S.
,
Reynolds
,
A. M.
and
Chittka
,
L.
(
2016
).
Life-long radar tracking of bumblebees
.
PLoS ONE
11
,
e0160333
.
Wystrach
,
A.
,
Beugnon
,
G.
and
Cheng
,
K.
(
2011
).
Landmarks or panoramas: what do navigating ants attend to for guidance?
Front. Zool.
8
,
21
.
Wystrach
,
A.
,
Mangan
,
M.
and
Webb
,
B.
(
2015
).
Optimal cue integration in ants
.
Proc. R. Soc. B
282
.
Wystrach
,
A.
,
Schwarz
,
S.
,
Graham
,
P.
and
Cheng
,
K.
(
2019
).
Running paths to nowhere: repetition of routes shows how navigating ants modulate online the weights accorded to cues
.
Anim. Cogn.
22
,
213
-
222
.
Zar
,
J. H.
(
2006
).
Biostatistical Analysis
.
Pearson
.
Zeil
,
J.
(
2012
).
Visual homing: an insect perspective
.
Curr. Opin. Neurobiol.
22
,
285
-
293
.
Zhang
,
S.
,
Bartsch
,
K.
and
Srinivasan
,
M.
(
1996
).
Maze learning by honeybees
.
Neurobiol. Learn. Mem.
66
,
267
-
282
.
Zhang
,
S.
,
Si
,
A.
and
Pahl
,
M.
(
2012
).
Visually guided decision making in foraging honeybees
.
Front. Neurosci.
6
,
88
.

Competing interests

The authors declare no competing or financial interests.

Supplementary information