the transfer of information in the dance language of honeybees: progress and problems

J Comp Physiol A (1993) 173:135-141 dminnal of

N~alll , a n d

�9 Springer-Verlag 1993

| l

i;

THE KING SOLOMON LECTURES IN NEUROETHOLOGY

THE HEBREW UNIVERSITY OF JERUSALEM

I I I I b

UI_

The transfer of information in the dance language of honeybees: progress and problems A. Michelsen

Institute of Biology, Odense University, DK-5230 Odense M, Denmark

Accepted: 15 April 1993

Abstract. The dance language of honeybees will always be associated with the name of Karl von Frisch, who was one of the two founders of Zeitschrift ffir Vergleichende Physiologie, now the Journal of Comparative Physiol- ogy. The discovery of the dance language has already led to a great number of investigations of physiological mechanisms, and more studies can be expected in the future. It therefore seems most appropriate to let this King Solomon Lecture deal with the progress and problems in our efforts to understand the transfer of information in the dance language of honeybees.

Introduction

Successful forager honeybees (Apis mellifera) are able to recruit other bees to a food source. For centuries, it was believed that the foragers lead the recruits to the food. Since the pioneering work of Karl von Frisch (review 1965) it has been known that the bees are recruited by odours and by dances, and that several components of the waggle dance are correlated with the direction and distance to food (or resin, water, or a new nest site). It has not been known, however, which of these components are perceived as signals by the follower bees, or how the follower bees could detect the dancer's movements in the darkness of the hive.

The dance language is an exception to the general rule that animals do not, on the whole, communicate about ideas or remote objects, but rather about immediate events connected with the actor and its surrounds (like readiness to mate or the approach of a predator). Fur- thermore, in the dance language an abstract, or symbolic,

code is used to transmit an impressive amount of information.

The waggle dance is named from the wagging run (in which the dancer wags her body from side to side and emits sounds by vibrating her wings). In the waggle dance, the dance path takes on a figure-of-eight shape: the dancer runs in a straight line (the wagging run) and circles back, alternating between a left and a right return path. The direction of the wagging run relative to the vertical (gravity) on the comb indicates the direction to the food relative to the sun's azimuth in the field. The velocity of the dance (and the number of figures-of-eight per unit of time) depends on the distance to the food source. A few follower bees keep close contact with the dancer, and these bees may be recruited to visit the food.

The main part of this review deals with two problems, which are probably interconnected: How can the follower bees observe the dance in the darkness of the hive? And which of the many components of the dance do the follower bees perceive as signals (i.e. which components are transmitting the information)? Comments are also made on the sensory mechanisms involved as well as on the coding and possible perception of information about location in round dances and profitability of food in both waggle dances and round dances. It is concluded that a reinvestigation of the biophysics and physiology of the air-flow-mechanoreceptors and further behavioural studies are needed.

How do the followers sense the dance?

Despite the work by Karl von Frisch and his colleagues it remained unknown how dance followers detect the

136 A. Michelsen: Information transfer in bee dance language

dancer's movements in the darkness of the hive. Karl von Frisch (1965) suggested two possible strategies: the dance sounds might travel through the wax comb from the dancer to the follower bees, or the followers might just touch the dancer. The first suggestion has been ruled out by experiments in which the vibrations of the wax comb close to dancing foragers were measured with a laser vibrometer while the air-borne components were mon- itored with a microphone. We found that the dance sounds travel exclusively through the air (Michelsen et al. 1986).

The second suggestion has not been ruled out, but it is considered very unlikely. Measurements on film re- cordings of the distances between the dancer and the bees surrounding it showed that less than 25% of the surrounding bees are close enough to be able to touch the dancer at any time (Michelsen et al. 1987). However, not all of the bees surrounding a dancer may be active followers trying to detect the dancer's movements, and the percentage of active followers able to touch the dancer is not known. In many cases there is no doubt that a follower briefly touches the dancer, but in other cases one cannot be sure (for example, the dancers tend to raise their wings well above the abdomen, and some follower bees may place the tips of their antennae between the wings and the abdomen). It may be argued that touching is not likely to be a good strategy for obtaining precise information about the position of the dancer, because the antennae are likely to be hit by the very violent wagging excursions of the dancer's body. Obviously, more work is needed on this question.

Measurements of the acoustic near field of dancing honeybees have suggested a third strategy (Michelsen et al. 1987). The near field is a zone close to a sound emitter, where the air particles oscillate with a much larger velocity than further away (and often in a complex spatial pattern), and where the sound pressure may also be appreciably higher. The dance sounds have a pressure amplitude of ca. 0.1 Pa (74 dB SPL) when measured at a distance of 1-2 cm from the bee (Michelsen et al. 1986). In contrast, the sound pressure measured with miniature probe microphones at the surfaces of the wings is ca. 1 Pa (94 dB SPL). In addition, pressure gradients up to 1 Pa/mm around the edges of the wings cause the air particles tO oscillate with velocities up to 1 m/s (peak- peak). The pressure gradients and air flows decrease with the third power of the distance from the dancer (Fig. 1), and the zone of intense air flows exists only close to the dancer's abdomen (close to the wings).

These observations and the fact that most follower bees are placing their heads in the zone of intense air flows led us to propose that the follower bees might extract the required information about direction and distance by detecting the air flows produced by the dancer. Further measurements have shown that an extremely complicated three-dimensional field of oscillating air flows exists close to the dancer. The air flows are generated not only by the vibrating wings, but also by the wagging movements (Michelsen et al., unpubl.). Some components of the wagging air flows may also have velocity amplitudes up to 1 m/s (peak-peak), but the

A~= O. 0:1

0 O" 0

t /

t t

/

/

/ ) 0.05 0.2

O" 0'0" - -

\

�9 0 21

"0 17

Fig. 1. Sound pressures (in Pa) measured in two directions radially away from a dancing bee by means of a pair of miniature microphone probes. Pairs of dots (bracketted to the left of the dancer) show the location of the two probes. Note that the pressure dif- ference (A p) decreases rapidly with distance from the dancer. In this figure, a A p of 0.2 (Pa over a distance of 2 mm) corresponds roughly to an air velocity of 0.1 m/s (peak-peak). Had the probes been oriented one above the other (not shown) and very close to the edge of the wings, air velocities of up to 1 m/s would have been recorded. The distance between the dancer and the follower bees in this drawing is the average distance between the dancer and the bees surrounding it. (From Michelsen et al. 1987)

physical nature of the wagging air flows differs from that of the air flows generated by the wings (by being deter- mined mainly by incompressible fluid mechanics and not by the laws of acoustics).

Amplitude spectra of the oscillating air flows generated by the wings are dominated by the frequency of wing vibration (250-300 Hz). In contrast, those of the wagging air flows tend to be much more complicated with a number of higher harmonics. For some components of the flow, the 10-15 Hz frequency of wagging does not contribute much. It is interesting that the spec- trum of the wagging air flow behind the dancer is very different from that lateral to the dancer. When bees are attracted to follow a dance, they generally approach the lateral side of the dancer, but gradually they then tend to move caudally. One may speculate that they are guided in part by the gradients in the wagging air flows. In addition, the wagging excursions produce an amplitude modulation of wing-generated air flows at the position of the follower bees, since the amplitude of these air flows decreases rapidly with the distance from the source. This amplitude modulation will not be experien- ced in the same manner by an observer (follower bee) lateral to the dancer and an observer behind the dancer (see Michelsen et al. 1987).

The sensory perception of oscillating air flows

Honeybees can be conditioned to respond to artificially generated oscillating air flows, which can also release a

A. Michelsen: Information transfer in bee dance language 137

wing-folding reaction (Kirchner et al. 1991). The thresholds for these responses are at fairly constant velocity in the frequency range from 10 to 400 Hz, above which they rapidly increase. If follower bees react in a similar manner to the air flows around the dancer, then they should not be able to perceive any higher harmonics of the air flows generated by the wings, whereas the higher harmonics of the wagging air flows might be perceived.

The thresholds observed in the study by Kirchner et al. were quite high (100-300 mm/s peak-peak). Although well below the maximum values for the two kinds of air flow around the dancer, this threshold would hardly allow the follower bees to detect the radial air flows at distances of more than a few mm from the wings (Fig. 1). The learning performance was poorer than in previous experiments which paired odour, colour or visual patterns with sucrose rewards. The authors argue that bees seem relatively unprepared to associate food with "sounds" (oscillating air flows). Consequently, these experiments do not rule out that the thresholds might be considerably lower in situations to which oscillating air flows "belong". The same argument can probably be used for the capacity for frequency discrimination. Kirchner and his colleagues found that bees trained to 265 Hz could later discriminate this frequency from 10 Hz and 380 Hz, but not from 110 Hz and 310 Hz. It would be interesting to perform similar experiments in which the bees were trained to the 10-15 Hz basic frequency component of the wagging.

In the study by Kirchner et al. the stimuli were the oscillating air flows at the open end of a tube of 11 mm internal diameter, connected at the other end to a loud- speaker. The length of the tube was chosen to give maximum oscillations at the open end. Unfortunately, the spatial properties of the oscillations were not described, but one may guess that the oscillations were fairly ho- mogeneous across the end of the tube and over small distances in the direction to the bee.

In contrast, the wing-generated air flows decrease in amplitude with the third power of distance from the source. The air flows caused by wagging show similar rapid decreases with distance (the exact relation depends on the nature of the component). It thus appears likely that the stimulus used by Kirchner et al. was rather different from the oscillating air flows around dancing bees. Furthermore, the pure 10 Hz air flows used by Kirchner et al. as a substitute for wagging air flows were very different from the natural stimulus. Obviously, in future studies the stimuli should be chosen in such a way that they reflect the complexity of the air flows that originate from a dancing bee.

This is especially important when the potential receptor organs are to be reinvestigated. The Johnston's organs in the antennae are the likely, but not necessarily the only, receiver systems for the oscillating air flows generated by dancing bees. The mechanics of the antenna was investigated by Heran (1959), who reported that the displacement amplitude of the antennal flagellum is at a maximum at ca. 240-280 Hz when the head of a bee is glued to the end of a vibrating rod. (This maximum is often referred to as a "resonance", but the maximum

disappears when the vibration amplitude is plotted as a velocity rather than as a displacement; it is therefore not a resonance in a physical sense). Heran also found that the threshold of Johnston's organ corresponds to a displacement of the antennal tip of ca. one Ixm. In these experiments the tip of the antenna was moved sinusoidal- ly. Obviously, the forces acting on the antennae during these experiments were very different from those acting on an antenna in the near field of a dancing bee.

A reinvestigation of the properties of the Johnston's organ is needed in order to determine its threshold, dy- namic range, and ability for neuronal coding when the antennae are subjected to controlled air oscillations of the type described here. In addition, we should study the properties of the various groups of mechanoreceptive hairs on the front of the bee's head, again using a semi- natural stimulus. The central processing of the information from these mechanoreceptors may provide further insight into the bees' ability to detect and analyze the complex air flows close to dancing bees.

A mechanical model of a dancing bee

We may be able to tell our bees to fertilize those apple trees five minutes fly to the south-east. To do this we should presumably need a model bee to make the right movements, and perhaps the right noise and smell. It would probably not be a paying proposition, but there is no reason to regard it as an impossible one.

J.B.S. Haldane: The Future of Biology (1927).

Since 1957, several unsuccessful attempts have been made to recruit bees by means of mechanical models of dancing bees (for references, see Michelsen et al. 1992). The first models did not emit sounds, but later models (e.g., wax-coated microphones or paralyzed bees) were wagging and emitting sounds. In all cases, the bees showed great interest in the models, but the models did not elicit recruitment. Our model (Fig. 2) differs from the previous ones principly by producing air flows similar to those observed close to dancing bees, and this model does recruit a number of bees to the locations indicated by its dances.

Our model was made of brass and covered with a thin layer of beeswax. It had the same length as a worker honeybee, but was somewhat broader. The wings were simulated by a single piece of razorblade, which was vibrated by an electromagnetic driver. A stepper motor rotated the model and caused it to waggle during the wagging runs. Other motors moved the model in a figure- of-eight path. During brief stops, a second stepper motor pumped "food samples" (scented sugar water) through a tiny plastic tube terminating near the "head" of the model.

All the motors and drivers were interfaced to and controlled by a computer, and all the dance components could be changed through the software. This made it possible to vary the individual dance components in- dependently and to create dances different from the normal waggle dance. At three-minute intervals the computer calculated the sun's azimuth and adjusted the direction of the wagging run to compensate for the (apparent)


Fig. 2. The mechanical model surrounded by follower bees during the performance of a wagging run. (From Michelsen et al. 1992)

A B

C D

E F

G

Fig. 3. Seven dance patterns tested in the experiments with the model dancing bee. Wagging and sound emission by the vibrating wing are indicated by a zigzag line and by a series of dots, respectively. A A normal wagging dance. B and C Dances with two and one displaced wagging runs, respectively. The wagging run(s) and the dance path indicate opposite directions. D Wagging of long duration combined with sound emission of short duration. E The opposite of D. F A dance with wagging during the wagging run and sound emission during a part of the return run. G The opposite of F. (From Michelsen et al. 1992)

movement of the sun. The model was deployed on the lower comb close to the entrance of a two-comb observa- tion hive. The sugar water and the wax coating of the model were given a faint floral scent.

Baits were placed at various locations in the field. In the bait, a piece of filter paper with 20 ~tl of pure floral oil placed below a metal mesh was renewed once every hour. At each location an observer noted the number of honeybees approaching the bait and showing the behaviour typical of a bee searching for a scented target. The bees did not receive food at the baits (thus we were sure to avoid dances by returning bees).

In the experiments on the transfer of distance information, seven baits were placed at various distances in the direction indicated by the model. In the normal dances (Fig. 3A) we selected dance parameters that in- structed the bees to fly for either a short distance (250 m) or a long distance (1500 m). When using the manipulated dances, we gave the bees conflicting information about whether to fly for a short or a long distance. In the experiments on direction, 8 baits were placed in different directions 370 m from the hive. Again, normal dances provided the bees with one direction, and manipulated dances indicated two potentially conflicting directions.

The experiments demonstrated that the follower bees are able to perceive information regarding both distance and direction f rom the wagging dances performed by the model bee. Furthermore, the bees were recruited not only by imitations of normal wagging dances, but also by some manipulated dances in which components of the dance were changed or shifted to other parts of the dance path.

The transfer of information about direction is shown in Fig. 4. Two experiments with normal dances were performed on consecutive days with the same wind direction, but with the target directions differing by 90 ~ . Ap- proximately 80% of the observed bees were found in the direction indicated by the model, This accuracy is similar to that obtained in experiments with live dancers. The number of bees observed at the 7 other baits is in agree- ment with the numbers observed at a similar distance in control experiments, in which the bees were fed scented sugar water f rom a motionless model bee.

The results obtained for the transfer of information about distance were less clear. Most bees searched close to the hive when the bees had been "told" to fly 250 m with normal dances. A higher percentage of the bees were observed far f rom the hive when the model indicated a


24

[] N=22

18 [] N = 33 Fig. 4. Two experiments with normal wagging dances testing the transfer of directional information on two consecutive days. The direction indicated by the model is shown with an arrow. The directions indicated by the model on each day are drawn as though they both pointed in the same direction, even though different directions were actually used on the two days. The number of bees observed at each of 8 baits located 370 m from the hive are indicated. The direction of the wind is indicated for each experi- ment. (From Michelsen et al. 1992)

distance of 1500 m, but again most bees searched close to the hive. Nevertheless, the distributions of observed bees at the baits changed significantly when the bees were "told" to fly to another distance.

Obviously, one or more cues for distance were lacking in the model's behaviour. It has recently been reported that the wing vibration frequency is ca. 315 Hz when dancing bees indicate a distance of 50 m, but only 207 Hz at 1600 m (Spangler 1991). It is possible, therefore, that the follower bees in our experiments interpreted the constant frequency of the wing vibrations (280 Hz) to indicate a fairly short distance, and that they became con- fused when the model indicated a distance of 1500 m by the velocity of its dance. Further studies are obviously needed on this problem.

The experiments with manipulated dances included a pair of dances designed to determine whether the bees obtain information about distance from the duration of the wagging run or the duration of the return run. In one type of dance, the duration of the wagging run corresponded to 250 m, whereas the duration of the return run corresponded to 1500 m. Another type of dance had the opposite design: the duration of the wagging run signalled 1500 m and the return run 250 m. In both cases, the bees followed the instructions given by the wagging run and ignored the duration of the return run.

The wagging run thus appeared to be the "master component" for conveying information about distance. That it also indicates direction was demonstrated by means of dance pattern C in Fig. 3, where the wagging run is displaced to one of the return runs. The bees were now provided with conflicting information about direction: the figure-of-eight dance path pointed in one direction and the wagging run in the opposite direction. In the direction indicated by the wagging run we observed ap- proximately one half of the bees, and less than one tenth

of the bees were observed in the direction indicated by the figure-of-eight dance path.

In contrast, dance figure B (Fig. 3) did not lead to any recruitment of the bees. The follower bees were obviously unable to predict where the next wagging run would be taking place. They were attracted by the wagging runs, but generally arrived too late to be able to follow the wagging run. They then remained on the spot, apparently waiting for the next wagging run - only to find out that it now occurred some distance away. They then ran to the place of action, arrived too late, and so on. One may speculate that the stereotyped figure-of-eight dance path may help the bees to orientate themselves relative to the dancer so that they are in a favourable position for observing the next wagging run.

During a normal wagging run the dancer both wags its body and emits sounds by vibrating its wings. In preliminary experiments (Michelsen et al. 1989) it was shown that both wagging and sound are necessary for the recruitment of the follower bees. We had expected that these components might have somewhat different roles in the transfer of information. For example, the distance might be signalled by the duration of one of them, and the other component might attract or motivate the bees.

This expectation was not supported by the results of several experiments, in which the wagging and wing vibration were partly or totally separated (dance figures D-G in Fig. 3). D and E were designed to test the roles of sound and wagging in the transfer of information about distance. One of these components was present during the entire wagging run (the duration of which corresponded to a distance of 1500 m). The duration of the other component was reduced to one third (which happened to correspond to a distance of 250 m). The results showed that although only one component was present during the entire wagging run, the bees still used the duration of the entire wagging run for estimating the distance to the target.

We then completely separated the wagging and sound components (F and G) and tested the transfer of directional information. Much to our surprise, in both cases the bees were recruited to the direction of the return run ! The scatter was very large, and in some tests the bees were not recruited at all, but the trend was clear (and significant).

The results of the experiments with dance figures D-G thus did not support the notion of different roles for the sound and wagging components. In contrast, these dance components seemed to be fairly redundant. Signal redundancy is known to make communication systems more tolerant to transmission errors (see MacKay 1972). Of course, the signal redundancy demonstrated here does not rule out the possibility that the sound and wagging may have additional and separate signal functions.

The number of bees recruited by our model was generally smaller than that observed with live dancers. Given the crude nature of our model, it is perhaps not surpris- ing that it induces only moderate recruitment. Neverthe- less, many applications would become easier if the model could be improved. Live dancers are known to have a thoracic surface temperature above 40 ~ that is 5-6 ~


above that of most other bees in the hive (Stabentheiner et al. 1990). We have tested a model with built-in heating elements. Unfortunately, it released fierce attacks by the bees each time its temperature exceeded 34 ~ The reason for this is not known. Here again, further work is required.

Further aspects and more unsolved problems

Our results with the mechanical model confirm that, as suggested by Karl von Frisch, the abstract information carried by the wagging dances is communicated to recruits. According to the odour theory of Wenner (1967) and his colleagues, the recruitment to specific sites should be due only to odour cues left in the field or carried by foragers. Such (hypothetical) cues were absent in our experiments, where the "forager" (the model) had not been in the field before dancing. In experiments like those shown in Fig. 4, the searching pattern of the recruits changed in accordance with the instructions received from the model and was independent of the direction of the wind, contrary to the expectations of the Wenner- theory.

The theory of Wenner and his colleagues was subjected to a critical examination and experimental test by Gould (1976), but it has proven more resistant to scientific argumentation than anticipated (Wenner and Wells 1990, reviewed by Gould 1992). Some philosophers and science sociologists, who uncritically accept the views of Wenner and his colleagues, have greeted the "dance language controversy" as a welcome example of what they see as severe limitations of the scientific method (e.g., Veldinck 1989). F rom a scientific point of view, the debate should have ended with Gould's review (1976), and any surviving skepticism should be buried after the experiments with the model bee.

Waggle dances are not the only dances for com- municating the location of food. In the round dance the dancer circles one or more times in one direction, then turns and circles in the opposite direction, and so on. Karl von Frisch (1965) interpreted round dances as re- cruiting signals for food sources within 50-100 m distance from the hive. He further assumed that round dances do not contain any precise information about the location of food sources, but this assumption was not correct. The round dances do indeed contain short wagging runs, which carry information about direction and distance, precisely as in waggle dances (Kirchner et al. 1988). For waggle dances, we have found that the wagging run is the master parameter for indicating distance and direction. It is tempting to assume that the bees perceive the messages in the round dance and the waggle dance in similar ways, and that they only pay attention to the wagging run, not to the entire dance path. We are, however, still lacking definite evidence for a behavioural role of the specific information carried by the round dance. With the present construction of the model bee, it is not possible to mimic round dances, so an approach to this problem will require some modifications of the model.

A similar situation exists for the possible transmission of information about the profitability of the food through the dances. Karl von Frisch (1965) described dances signalling profitable sources as more "lively". Several investigators report a correlation between profitability and a number of the components of both waggle dances and round dances (e.g., Esch 1962; Wad- dington and Kirchner 1992), but so far it has not been demonstrated that the follower bees perceive this information and use it when deciding whether they should visit a source. Here again, a modified version of the model dancer may prove useful.

In this review, I have concentrated on the behaviour, signals and possible sensory mechanisms. The hypothesis that follower bees may perceive the communication dances through the oscillating air flows surrounding the dancers is now a likely explanation, but it cannot be regarded as a proven fact. The guesses about the receptor organs involved also have to be substantiated. Even when we reach a reasonable understanding of these ques- tions, the central nervous processing of all this complicated information still remains as a major challenge.

Acknowledgements. Most of the work discussed in this review is the result of a close cooperation over several years between two very different research groups, specialized in biophysics and behaviour, respectively. I should like to use this opportunity to thank Professor Martin Lindauer, Wiirzburg University, for his friendship and for our excellent cooperation. Without his profound understanding of honeybees, our modem technology would not have been of much use. Our studies have been generously supported by grants from the Danish Science Research Council, the Carlsberg Foundation, the Akademie der Wissenschaften und der Literatur (Mainz), the Hum- boldt Stiftung, and the Stiftung Volkswagenwerk. I am most grate- ful to Professor Jeffrey Camhi for inviting me to present this lecture in Jerusalem and to Martin Lindauer and Ole N~esbye Larsen for comments on the manuscript.

References

Esch H (1962) Uber die Auswirkung der Futterplatzqualit~it auf die Schallerzeugung im Werbetanz der Honigbiene. Verh Dtsch Zool Ges: 302-309

Frisch K von (1965) Tanzsprache und Orientierung der Bienen. Springer, Berlin Heidelberg New York

Gould JL (1976) The dance-language controversy. Q Rev Biol 51:211-244

Gould JL (1992) Interpreting the honeybee's dances (review of Wenner and Wells 1990). Am Sci 80:278-279

Heran H (1959) Wahrnehmung und Regelung der Flugeigen- geschwindigkeit bei Apis mellifica L. Z Vergl Physiol 42:103-163

Kirchner WH, Lindauer M, Michelsen A (1988) Honeybee dance communication: Acoustical indication of direction in round dances. Naturwissenschaften 75 : 629-630

Kirchner WH, Dreller C, Towne WF (1991) Hearing in honeybees: operant conditioning and spontaneous reactions to airborne sound. J Comp Physiol A 168 : 85-89

MacKay DM (1972) Formal analysis of communicative processes. In: Hinde RA (ed) Non-verbal communication. Cambridge Univ Press, Cambridge, pp 3-25

Michelsen A, Kirchner WH, Lindauer M (1986) Sound and vibra- tional signals in the dance language of the honeybee, Apis mellifera. Behav Ecol Sociobiol 18:207-212


Michelsen A, Towne WF, Kirchner WH, Kryger P (1987) T h e acoustic near field of a dancing honey bee. J Comp Physiol A 161 : 633-643

Michelsen A, Andersen BB, Kirchner WH, Lindauer M (1989) Honeybees can be recruited by means of a mechanical model of a dancing bee. Naturwissenschaften 76:277-280

Michelsen A, Andersen BB, Storm J, Kirchner WH, Lindauer M (1992) How honeybees perceive communication dances, studied by means of a mechanical model. Behav Ecol Sociobiol 30:143-150

Spangler HG (1991) Do honey bees encode distance information into the wing vibrations of the waggle dance ? J Insect Behav 4:15-20

Stabentheiner A, Hagmiiller K, Kovac H (1990) Thermisches Ver- halten von Honigbienen im Schw~inzeltanz. Verh Dtsch Zool Ges 83: 624

Veldinck C (1989) The honey-bee language controversy. Inter- discipl Science Rev 14:166-175

Waddington KD, Kirchner WH (1992) Acoustical and behavioral correlates of profitability of food sources in honey bee round dances. Ethology 92:1-6

Wenner AM (1967) Honey bees: do they use the distance information contained in their dance manoeuver? Science 155:847-849

Wenner AM, Wells PH (1990) Anatomy of a controversy: The question of "language" among bees. Columbia University Press

the transfer of information in the dance language of honeybees: progress and problems

Documents