ANATOMY OF A CONTROVERSY: THE QUESTION OF A “LANGUAGE” AMONG BEES (Columbia University Press, 1990) pgs: 111-128
ADRIAN M. WENNER
PATRICK H. WELLS
“The brain of a bee is the size of a grass seed and is not made for thinking. The actions of bees are mainly governed by instinct. Therefore the student of even so complicated and purposeful an activity as the communication dance must remember that he is dealing with innate patterns, impressed on the nervous system of the insects over the immense reaches of time of their phylogenetic development.”
-Karl von Frisch 1962:78
“Anomalies . . . are the commonest intellectual vehicles for breaking through; all are solvable in the sense that any one is understandable, but that one leads with the power n to still more and deeper anomalies.”
-John Steinbeck (1941) 1962:150
Anomalies encountered during scientific research, although most often ignored due to paradigm hold, at other times provide a basis for solid research. However, even though anomalies are common and valuable during the conduct of research, anomalies and their impact can quickly fade from consciousness as more solid evidence gains priority; what is anomalous under one paradigm becomes the expected after a “conversion” to another paradigm (gestalt switch).
In this and in the following chapter, we include anecdotes from our own experience as examples of the importance of anomaly and how it directed the course of our own research. The recognition of the existence of those anomalies and the consideration of various interpretations led to changes in our subsequent experimental designs. Augmentation of our own experiences by the input of even a single anecdote from other people often helped create new images in our minds.
Edward Jenner’s account of studies of the link between cowpox and smallpox is a significant example of what would now be considered an unacceptable inclusion of anecdotes in a published paper. That research, however, led to the eventual worldwide eradication of smallpox (see excursus JNR). Jenner included a series of anecdotes in order to augment his case, but (as indicated above) a single anecdote may help “create an image” and lead one to new interpretation.
“LEARNING” VERSUS “INSTINCT” AMONG INSECTS
In the 1940s and 1950s only a few students of insect behavior explained behavioral patterns in terms of learning. Most others proceeded strictly within an “instinct” paradigm while interpreting behavioral patterns (see the von Frisch epigraph, above). The “spirit of the times” was permissive; either explanation could be used for a given behavioral act, despite the lack of experimental tests.
Earlier, Loeb ( 1973) and Fraenkel and Gunn ( 1961) had advocated the use of a more quantitative approach to descriptions of animal behavior, including a series of terms that persist today (e.g., “taxes” and “kineses” for animals and “tropisms” for plants). Since 1950, however, the prevailing thought in ethology has shifted gradually toward ever less well-defined terminology and assumptions. The “instinct” versus “learning” issue remains obscure and largely ignored in the animal behavior community (Rosin 1980a, 1980b; see also our chapter 13).
THE “INSTINCTIVE LANGUAGE” OF BEES
Research on honey bee communication during the 1940s and 1950s, then, still relied heavily upon the notion that the behavior of these small animals was largely “innate” or “instinctive” (see the von Frisch epigraph). Thorpe summarized that attitude as follows:
Much of the recent work on insects and other arthropods seems to fit in very well with the hierarchy concept of instinctive behaviour of Tinbergen. Perhaps this fact . . . receives some further elucidation from recent researches on the mode of action of the insect nervous system. Thus Vowles (1961) points out that the properties of the insect neurone, which are very different from that of the vertebrates, and the small size of the insect nervous system, render necessary a functional organisation of behaviour far simpler than is often supposed. (1963:231)
The attitude that insect behavior was fundamentally more simple than that of vertebrates would seem to exclude the possibility of a “language” among bees (Rosin 1978). However, if one pursues research within the verification approach (the Carnap arm of the Realism school), it is possible to retain both notions; a simple system can permit the existence of “language” if one considers the more complex behavior to be merely a fixed sequence of simple steps. Tinbergen verbalized that amalgamation as follows:
When “unemployed” honey bees, waiting in the hive for a messenger, are at last activated by one performing the “honey dance,” . . . the stimulus delivered by the dancer bee stimulates them to leave the hive. They fly in a definite direction over a definite distance (both communicated to them by the dancer) and begin to search for flowers, selecting only those that emanate the scent carried by the messenger. They suck honey [sic], and after having made a “locality study,” they fly home. In this latter case the stimulus given by the messenger [dancing bee] releases a complicated behaviour pattern. (1951:54, 55)
The “chain reflex” explanation of behavior, as used by Tinbergen and others at that time, permitted the phenomenon of honey bee recruitment to food sources to be known eventually as an “instinctual signaling system.” The teleological notion of “purposefulness,” which later came to be so fundamental in ecology and sociobiology, also was relied upon in those early days. Thorpe, for example, invoked that concept in describing honey bee recruitment to food sources, as follows:
An insect with such a high degree of organisation of labour, and having only a limited period of the year in which to forage, needs some means of communication by which a scout which has discovered a rich source of food can quickly recruit a body of workers large enough to fetch the available booty but not so great as to waste the worker strength of the hive in unprofitable foraging. . . . In other words, bees need a language; and von Frisch . . . did not hesitate in his earlier papers to speak of “the language of bees.” (1963:268)
The above summary provides only a sketch of the attitude prevalent in insect behavior research when we began work in the late 1950s. However, at that time we agreed completely with both the notion that bees were bound to simple behavioral patterns (i.e., that they exhibited “instinctive” behavior) and to the notion that honey bees had a ‘language.” In that sense we accepted the essence of Tinbergen’s above expression, that the “dance language” of bees was little more than a fixed sequence of simple behavioral acts (“chain reflex” behavior). That positive attitude was possible because we had also used the same verification approach (the Realism school in the Carnap sense) in our research that other researchers in animal behavior had used in theirs.
The first several years of our research were thus conducted well within the limitations of the verification approach; our goal was to verify (“prove”) the existence of a language already “known” to exist. We concurred wholeheartedly with von Frisch’s statement: “The language of bees is truly perfect, and their method of indicating the direction of food sources is one of the most remarkable mysteries of their complex social organization” (1950:75).
A Verification Approach: Sound Waves as a Language Element?
The first nine-year period (1957-1966) of our research was thus conducted under the premise that honey bees had a “language” and that they behaved as outlined in the passages quoted above. There was no reason to suspect otherwise at the time. However, as indicated above and in chapter 2, it was already apparent to us by 1965 that the von Frisch dance language hypothesis, appealing as it was, rested entirely on circumstantial evidence. Even though we worked within the verification approach, we felt that a more direct type of supportive evidence was needed (as opposed to the existing circumstantial evidence).
By the mid-1960s, then, several years of research had been spent pursuing the possibility that bees somehow made direct use of the sounds they made during their waggle dance (e.g., Wenner 1959, 1962, 1964; Esch 1961), “while communicating the location” of food sources. The discovery that sounds existed within the dance maneuver had provided an entirely new set of possibilities regarding communication among bees (Wenner 1964).
Admittedly, this research activity was not altogether a detached scientific pursuit (Wenner 1971a); “discovery” and “proof” are essential elements leading to success within any scientific community, particularly when it functions primarily within the verification approach. Encouragement at that time came from all quarters; other scientists were quite receptive to the possibility that honey bees were perhaps capable of an even more complex behavior (anthropomorphically, an “acoustic speech”) than had been reported earlier.
The spectrographic pattern obtained from an analysis of sounds produced by dancing bees (see figure 7.1) differed markedly from sounds made by bees engaged in other behavioral acts (Wenner 1964). Most bee sounds are continuous tones, but during each straight-run portion of the waggle dance (see figure 7.1) foragers produce a burst of pulsed sound (amplitude modulation). The duration of that train of pulses is correlated with the distance a forager has traveled between hive and feeding place in the field (see figure 7.2).
It is tempting to examine the foregoing correlation in a teleological context. One could argue that a correlation between sounds produced in the hive and distance traveled in the field would not exist unless it had a purpose. One could also argue that amplitude modulation in a “sound signal” also would not exist unless it were somehow “useful” to recruit bees. Phrased teleologically: “Bees would not waste that much energy in a purposeless act” (see excursus TE).
Moreover, the location in the hive where the dance is conducted is often exceedingly dark. That is because the hive entrance is normally very small, and there are a great number of bees with an abundance of hair on their bodies between the entrance and surface of the comb; both of these factors would restrict the amount of light that could penetrate in as far as the dance area on the combs.
There is also the problem of dance attendants and their orientation on the comb relative to the position of the dancing bee. Virtually all early diagrams and their legends indicated that potential recruits “followed” or “tripped after” the dancing bee (e.g., figure 46 in von Frisch 1967a). However, we have noticed that recruits are most often positioned at right angles to the dancer during the straight-run portion of the dance (see figure 6.11).
The antennae, which contain sound-sensing units, of those recruits are also in intimate contact with the body of the dancer when sound is being produced (see figure 7.1). Under all of the above circumstances, one could reasonably conclude that sound signals were an important constituent of the “dance language” of bees.
While von Frisch was “decoding the dance language of bees,” he was unaware that dancing bees made those peculiar sounds during the straight-run portion of the waggle dance. He therefore had researched only some of the possibilities by which communication of food location could occur, if indeed bees had a “language.” Several years later von Frisch recognized the potential importance of sounds during the waggle dance when he wrote: “Taking all . . . facts into account, we regard the acoustically emphasized duration of waggling as the index of distance” (1967a:104).
The discovery of sound production during the dance represented to us the first of many anomalies that arose during our research program. More important to us (in retrospect) was the fact that we began to become more aware of the multiplicity of explanations and approaches that could be important during any investigation of a problem in science.
Conditioned Responses: Another Anomaly
A major and drastic turning point in our research occurred when it abruptly became apparent to us that bees could learn very rapidly in the classic conditioning sense. Our surprise (gestalt switch) was due to the fact that most of the earlier von Frisch results (1947, 1950) that he had submitted as evidence of “language” use could be interpreted instead as merely the result of conditioned responses to stimuli (Johnson and Wenner 1966). One then no longer needed to postulate an elaborate “instinctual signaling system” to explain the von Frisch results – a simple conditioned-response explanation would suffice.
It had long been known that honey bees can be “trained” to visit a feeding station (e.g., Maeterlinck 1901; von Frisch 1950). However, prior to 1965 that particular behavioral pattern had not been placed within the context of various theories of learning behavior. That training phenomenon is now recognized as a form of “choice discrimination conditioning” or “simple conditioning” (Wenner and Johnson 1966); bees can be trained to visit blue rather than yellow flowers or to visit a square design rather than a circle (see Wells 1973).
Each bee will, in fact, essentially train itself upon first visit to a colored and/or scented food source. From then on it will be a constant forager on sources of that type, as long as sufficient reward remains available (Wells, Wells, and Smith 1983; Wells, Wells, and Contreras 1986). Neither these well-known abilities nor the interpretation of their roles in behavior have ever been challenged.
Evidence that insects could also perform “simple discrimination conditioning” was quite inadequate before the 1960s (see Thorpe 1963 for a complete review of literature up to that time). This type of conditioning involves a pairing of some apparently “neutral” stimulus with the presentation of a reward. Later, when one presents that same stimulus in the absence of a reward, the animal nevertheless proceeds to behave as if a reward were imminent. One of the more famous examples is that of Pavlov’s dog being conditioned to salivate at the ringing of a bell.
We found quite accidentally that honey bees could learn in the classical “simple discrimination conditioning” sense (Wenner and Johnson 1966; Wenner 1971a). That incident occurred during a “normal science” (e.g., Kuhn 1962) sequence of experiments. Inadvertently, during laboratory studies, a “reward” of sugar solution had been provided each time a neutral stimulus (a draft of air) had been administered. Later, when an air draft was provided while the reward was momentarily delayed, the experienced bees rushed out to the empty food dish as if the sugar solution had already been provided.
When the bees thereby demonstrated that they could learn in the classical conditioning sense, the “My God!” reaction of Bruner and Postman (1949) followed immediately. First there was disbelief. Then there emerged a strong desire to dismantle the apparatus and tell no one about this apparent anomaly. However, the urge to learn more about what bees really do in nature prevailed over that temptation.
We had first learned of the Bruner and Postman experiments from Kuhn, who wrote:
In a psychological experiment that deserves to be far better known outside the trade, Bruner and Postman asked experimental subjects to identify . . . a series of playing cards. Many of the cards were normal, but some were made anomalous, e.g., a red six of spades and a black four of hearts. . . . After each exposure the subject was asked what he had seen. ( 1970a:62-64)
These psychologists had thus altered some of the playing cards in an otherwise standard deck; the new reality did not conform with any earlier experience the subjects had had with playing cards. When subjects first viewed altered cards, they initially did not recognize them as other than normal and continued to record a number and/or color they expected to see. They consequently erred whenever one of the altered cards was shown to them. Only when an inordinate amount of time was allowed for viewing each false card did they begin to perceive that their earlier identifications had been in error. A common expression among these experimental subjects when they first realized the fact that they had earlier erred was “My God!”
The experience we had when bees clearly demonstrated that they were capable of learning in the classic manner paralleled the playing card experience studied by Bruner and Postman. All common preconceived notions, such as the instinct-versus-learning distinction between insects and vertebrates, were suddenly open to question. The immediate reaction of shock when bees demonstrated their ability to learn was soon replaced by a need for us to return to reality. It became evident that it was no long possible to ignore either what had transpired or what adverse reactions might be ahead in the scientific community (see chapter 12).
We had thereby been propelled from an exercise in normal science (Kuhn 1962) into the “image creation” experience of Atkinson (1985). Unwittingly we proceeded thereafter under the mistaken belief (see chapter 9) that the results of more tightly controlled experiments would be welcomed by the same community that had welcomed our research on analysis of honey bee sounds. We also felt that we could succeed in “converting” others to the important role of learning in honey bee recruitment.
However, studies of learning remained a side issue. We still believed that honey bees had a “dance language,” albeit one of reduced importance in foraging behavior, that functioned during the first recruitment of naive bees to a food source. After several experiments on simple conditioning, we again returned to the question of what might constitute “vigor” in the waggle dance in order to attempt to construct an imitation bee (see excursus VGR). That approach would permit us to gather “direct” evidence in support of the notion that sound signals were an essential component of dance language (the verification approach once again).
While our intent was clear, the approach instead led (inadvertently at first and deliberately later) to a sequence of experiments that further clarified the role of learning in the recruitment of honey bees to food sources in the field (e.g., Johnson and Wenner 1966).
The first stage in subsequent research included experiments that actually extended our understanding of the “simple conditioning” phenomenon. We first confirmed, as others had found earlier (e.g., von Frisch 1950), that foragers accustomed to visiting a station at which they have previously been successful will continue to visit that site, even if food is no longer available. In fact, their visits, on average, are remarkably regular (see figure 7.3). That result agreed fully with von Frisch’s perception:
We then stopped supplying sugar at both places, and allowed the dishes to remain empty for an hour or two. After this time most of the bees from both groups were sitting inactive in the hive; only from time to time would one of them fly out to the feeding place to see if anything was to be had. (1950:72)
However, neither observation agreed with a later statement by von Frisch:
Suppose we remove the little sugar-water dish from our feeding table, so that our marked bees find that there is no food in the usual place? They will behave exactly as they would if their natural food, the honey flow, had dried up owing to bad weather, when their usual flowers temporarily cease to provide them with nectar. The bees will stay at home, and stop dancing. From then on the little honey dishes laid out round the hive may have to wait on the lawn for hours or even days on end before a single bee will visit them again. (1954:105)
Nor with yet another of his comments:
There still remains one factor that plays a part in the frequenting of flowers by bees: their pronounced time sense. . . . I know of no other living creature that learns so easily as the bee when, according to its “internal clock,” to come to the table. . . . These relations are easily imitated experimentally. If at an artificial feeding station one offers sugar water at a set time of day, within a day or two the visitors adjust themselves to the schedule. Thenceforth they come at the designated time, whereas before and after the hour of feeding even informational flights are almost entirely omitted. The foragers remain sitting at home, saving their strength and risking no unnecessary flights. (1967a:253-253)
Von Frisch apparently did not realize that his statements were not consistent with one another (the appearance of anomaly, once again).
Peter Craig at the University of California, Santa Barbara (unpublished results), repeated a 1929 experiment performed by Beling (figure 35 in Ribbands 1953). He had thirty-five individual bees trained to visit a feeding station at which food had been provided for several days only between 4 and 5 p.m. Craig then tallied all visits by each of the thirty-five bees; he found that some of them inspected the dish more than once during the day. That was the same result found by others (summarized in chapter 7 of Ribbands 1953).
Craig then recognized a problem with data display in earlier studies. The repeated tallying of the same bee visiting a station provided an impression of greater precision in “time sense” than was merited by the results. That is, a forager in the general area of the feeding dish could periodically reinspect the dish without returning to the hive. Each such visit was counted as an additional point in Beling’s display.
Figure 7.4 presents Beling’s results in a different manner. The data are now included for only the first visit of the day by each of the foragers. By the beginning of the training period (2:30 p.m.) more than half of the foragers had already inspected the dish.
Craig’s data are shown in the same figure for comparison. By 4 p.m. (the start of the training time) 95 percent of the experienced bees in Craig’s study had already inspected the empty dish at least once during the day. More than half of them had done so at least an hour and a half early, and a tenth of them had already inspected the dish five hours before the training time.
In both experiments, it is apparent that foragers did not “remain sitting at home” (von Frisch 1967a:254). The “pronounced time sense” was certainly not impressive.
“LEARNING” AND THE EVIDENCE FOR “LANGUAGE”
The second stage of our experimental program was an attempt to repeat von Frisch’s experiments in order to clarify his comments about what he meant by differences in dance “vigor.” That is, he had indicated that bees visiting a rich food source dance more “vigorously” than those visiting a poorer food source.
It should be possible to elucidate this “vigor” effect by altering sugar concentration at a dish in the field while simultaneously observing the foragers upon their return as they danced in the hive. In conducting such an experiment, one must first observe dancing bees carefully to ascertain just what features of the dance might vary. During the first attempt at such a study, we used a hive and feeding station that were already in operation. One of us was to observe dancing bees in the hive while another person added a sugar solution of known concentration to the dish that had been empty since the previous day.
Initially all went as planned. Whereas the visitation of experienced foragers was linear and regular when no food was present in the dish (see figure 7.3), the cumulative number of arrivals became exponential once food was again provided (see figure 7.5). That result is exactly what one would expect if successful bees had recruited others. The result, however, was anomalous, in that an exponential rise in visitation occurred at the dish even though no dancing had occurred in the hive. Clearly, bees had been recruited by successful foragers despite the lack of dancing during that initial fifteen-minute period.
Many experiments later we could interpret those anomalous results. One can associate a stimulus, such as odor, with the reward to be provided. When that is done, one can later inject odor into the hive without providing a reward. We then did just that by injecting odor into the hive by using a turkey-basting syringe. Experienced bees then immediately left the hive and flew directly to the familiar, but empty, dish in the field (Johnson and Wenner 1966).
In nature the same condition can hold true. Experienced bees will occasionally inspect the empty food dish (or blossoms not yet producing nectar). Once foragers can again fill up, they will return to the hive and unload. While there, the odor emanating from their bodies can alert other experienced bees that food is again available. Recruitment can then occur even in the absence of dancing.
When that interpretation became evident from the experimental results, we passed through another “conversion” sequence, as described by Kuhn:
Initially, only the anticipated and usual are experienced even under circumstances where anomaly is later to be observed. Further acquaintance, however, does result in awareness of something wrong or does relate the effect to something that has gone wrong before. That awareness of anomaly opens a period in which conceptual categories are adjusted until the initially anomalous has become the anticipated. At this point the discovery has been completed. (1962:64)
The first sentence of that quotation is particularly significant with regard to the conditioned-response phenomenon we observed. Earlier workers must have observed the same “communication by means of conditioned response” behavior among recruited bees but had apparently failed to recognize its potential significance during the recruitment of experienced foragers to food sources. In that connection, consider a series of earlier quotations relevant to conditioning and recruitment.
One significant sentence can be found in the initial report of von Frisch’s experiments with honey bee recruitment. He wrote: “It follows further that a communication can be transmitted from the returning bee to other bees by touch alone, without the necessity for any dance” (1947:13). However, that qualification was absent from a later summary of his 1946 results:
The dances are apparently understood by the bees in the hive, as could be shown by the following experiment. . . . If we now refilled the dish at [a] distant site, then the wagging dances of the first gatherers to return with full stomachs aroused chiefly bees from the group which had previously visited the distant feeding place. But when we offered sugar-water at [a] nearer site, then the resulting round dances aroused mostly bees which had previously been feeding there. (1950:72)
On the basis of results from the experiments on conditioned responses described above, (e.g., Wenner and Johnson 1966; Johnson and Wenner 1966), it would appear that the von Frisch results, which supposedly had demonstrated the use of “language,” agreed equally well with the interpretation that conditioned bees had responded to some simple stimulus (odor?) provided by a returning bee.
Ribbands (1954) had also obtained results earlier than we, results that indicated that experienced foragers could be recruited to their food site by means of odor alone. He published a paper to that effect: “Communication between honeybees, part 1: The response of crop-attached bees to the scent of their crop.”
However, Ribbands apparently functioned within the paradigm hold described by Kuhn; that is, “only the anticipated and usual are experienced where anomaly is later to be observed.” In fact, a year later Ribbands published the second paper of the series, “The recruitment of trained bees, and their response to improvement of the crop.” An inspection of some of his data and one of his conclusions reveals that he came very close to recognizing what we published a little more than a decade later. He wrote:
In favourable conditions recruitment is very rapid. For instance, on the afternoon of 20th August dish b was put down at 13.20 hours G.M.T. and the first trained bee arrived at 13.36 hrs. Two others came at 13.38, and another one at 13.39 – the absence of bees [at another dish] indicates the probability that these bees were recruited by the first bee on its first return to the hive; it returned itself at 13.39. Another bee arrived at 13.41, two at 13.42, two at 13.43, four at 13.44, one at 13.45, one at 13.46, one at 13.48, one at 13.49, two at 13.51, one at 13.52, one at 13.54. One other did not come until 15.15 (perhaps alerted by vigorous dancing after the increase in syrup concentration at 14.00 hrs.). (1955a:27)
Just how close Ribbands came to our notion of “communication by means of conditioned response” is evident in one of his conclusions in that same paper (1955:31): “Lindauer reported that bees did not dance until they had paid several visits to a food source (at or near threshold syrup concentrations). The arrival times of the recruited trained bees . . . are only consistent with the supposition that the first arrival danced on her first return to the hive” (1955:31; emphasis ours).
The data published by Ribbands (above) escaped our attention at the time when we began our experiments. That was partly because we were unprepared for a conditioned-response behavior in bees and partly because Ribbands’ results were in the form of a paragraph rather than a figure. By expressing his data in the form of a graph (see figure 7.6), however, it is evident that our results were essentially identical to his. The main difference was that we were observing the foragers in the hive and knew that they were, indeed, not dancing upon their return (just as Lindauer had observed). The conditioned-response explanation was sufficient to explain our results, but Ribbands did not recognize the implications of his earlier statement: “the mere presence of the training scent in the hive, in the absence of either food sharing or dancing, can encourage crop-attached bees to go to their crop” (1954:143).
A concise statement of the alternative hypothesis for the recruitment of experienced bees was now possible.
Foragers routinely monitor known sources of food even after those sources become empty. (That statement agrees with one but not with another of von Frisch’s statements, above.) If food again becomes available, returning loaded bees enter their hive, bearing the characteristic odor of the food source and/or of the location on their bodies. Other foragers that have visited those same sources are stimulated to leave the hive by the odor stimulus carried in by the first successful forager(s) and travel to whatever site at which they had earlier had success (Johnson 1967b).
One very important fact had emerged from our studies of conditioned responses. In his classic little Cornell University Press book (1950) von Frisch made a number of claims with respect to what he felt was the conclusive nature of experiments purporting to demonstrate the use of “language” by bees. Yet the results of every one of those experiments could also be interpreted as an example of the behavior of bees during a conditioned response to an odor stimulus.
Increasingly we had come to appreciate also the importance of odor in the recruitment of naive honey bees to those food crops visited by their more experienced hivemates. That increasing awareness of ours is the subject of the next chapter.