10 June 2005
Once again we are treated to a presumed final proof of bee “language.” The latest effort (“The flight paths of honeybees recruited by the waggle dance”) appears as a letter on pages 205-207 in Nature on 12 May 2005 by J. R. Riley, U. Greggers, A. D. Smith, D. R. Reynolds, and R. Menzel. Is the question of bee language now resolved? Hardly, because the nearly four-decade old controversy does not revolve about evidence – as will be documented in this treatise.
Ruth Rosin recognized that element of the problem when she posted (6 June 2005) the following comment (in part) on the honey bee e-mail network (BEE-L):
“The Nature (2005) radar-tracking study by Riley et al. has already created considerable excitement in the popular scientific news-media. The study, nonetheless, never did, nor could, salvage the dance language hypothesis, because the whole study is simply totally irrelevant to the dance language controversy”
“According to the report only 2 bees, (out of the 19 radar-tracked bees released near the hive, for which flight-tracks are provided) “landed at the feeder.” In response to questions about various details concerning the study, Uwe Greggers (the scientist who actually designed the study, however, informed me, among others, (in e-mail exchanges), that those 2 bees did not actually land at the feeder. Instead, they only landed on the stand on which the feeder stood, but never found the food, or the feeder (even when they were not more than 8 cm away from the feeder). Anyone who questions that is free to personally check with him.
Two days earlier she had also written to me (in part):
“There is, however, another, much more basic problem with the Nature (2005) study. The whole study is simply totally irrelevant to the DL controversy. Before seeing the published report I assumed that the experimenters did what they should have done first and foremost, i.e. radar-track recruits recruited by foragers feeding on scented food; which is what foragers invariably do in nature. It turns out that the experimenters did exactly the opposite. They strove very hard, and apparently succeeded in radar-tracking only bees that never found any food.”
“The honeybee DL hypothesis was intended to provide an answer to the problem how honeybee recruits find their foragers’ food-source, and other sources with the same major odor, in the field. In no way is it possible to provide any answer to such a problem by studying only bees that never found any food during the study. Since the whole study is totally irrelevant to the problem, any results obtained in the study cannot be relevant to the problem, either.”
A Broader Problem
During the past 35 years, the dance language controversy has centered on the question, “Can someone prove a hypothesis true and expect that hypothesis to become a fact, a fact no longer open to question?” Bee language advocates would have us believe so. Those who study scientific process feel otherwise.
We published the results of our double controlled and strong inference experiments in the late 1960s and 1970s, results that did not mesh with predictions of the dance language hypothesis (see below). Ever since then, research on honey bee recruitment to food crops has remained in a crisis state, with millions having been spent on efforts, in various series of experiments, to “prove,” once more, that bee language is real. In each attempt, bee researchers have admitted that previous attempts had not sufficed – but did so only after they felt certain that their own efforts had succeeded.
The recent radar tracking studies of bee flight paths that left the hive is just one more in such a series of attempted “proofs.” On 14 October 2003 James Fischer posted the following statement on BEE-L, one made by those involved in that radar research:
“We have used harmonic radar to measure the flight trajectories of bees recruited after observing the waggle dance, this has enabled us to settle (hopefully once and for all) this controversy in favour of Von Frisch.”
In the first few words of their 2005 Nature report, the radar tracking researchers also used the phrase, “the dance language” of bees instead of more objective phrases, such as: “the waggle dance,” or “the dance maneuver.” It became clear soon thereafter in their letter that they had started their experiments with the assumption that bee language is real and then had tried to prove it true (circular reasoning?).
The renowned science philosopher Karl Popper would have been appalled at such a brazen and naïve approach. For instance, in 1957 he stated: “It is easy to obtain confirmations, or verifications, for nearly every theory – if we look for confirmations.” In scientific experimentation, the easiest one to convince is oneself. That is why we need blind, double controlled, and other rigorous designs in experimentation – elements lacking in the present study.
Karl von Frisch abandoned an earlier odor-search hypothesis and proposed his dance language hypothesis in 1946, a conclusion that rapidly became accepted nearly universally (including by me for my doctoral studies in the 1950s). That hypothesis provided a rational (though not exclusive) explanation about how newly recruited bees might manage to find a food source exploited by regular foragers. His interpretation morphed into the status of fact (“proved” or “discovered”), rather than remaining in people’s minds the hypothesis that it was.
What few appreciate today is that von Frisch’s language conclusion was at odds with his earlier published results and conclusion; that is, recruited bees use odor, and only odor, after leaving their parent colony in search of the crops exploited by experienced foragers. In 1937, for instance, he wrote:
“I succeeded with all kinds of flowers, with the exception of flowers without any scent. When the collecting bee alights on the scented flowers to suck up the food, the scent of the flower is taken up by its body-surface and hairs, and when it dances after homing the interested bees perceive the specific scent on its body and know what kind of scent must be sought”
For more information on that point, please access:
The exotic notion of bee “language,” though unproven and lacking rigorous experimental tests, eclipsed von Frisch’s earlier odor-search hypothesis. One prominent bee researcher, H. Kalmus, held out. He wrote in 1960 (Simple Experiments with Insects. Doubleday & Company, Inc. Garden City, NY. p. 96):
“The explanation [of how a forager communicates this information], however, is really quite simple; and any fairy tales about one bee telling the others, or leading the others to a locality, can be discounted. When a bee returns to the hive she starts dancing on the combs beating her wings and thus spreading the smell of the flower which clings to her body. The other bees become interested by the dance and go searching for that particular smell.”
That comment by Kalmus thus matched the firm conclusion von Frisch had published in 1937.
Although I conducted my doctoral research in the 1950s with the belief in bee language, later experiments by my co-workers and myself yielded results clearly at odds with that hypothesis. A series of unfortunate and unexpected events (for example, see Excursus SI from Anatomy of a Controversy) led us to execute rigorous experimental tests of the language hypothesis during the 1960s, the first such real tests ever conducted. (For clarification, experiments designed to gain support for a hypothesis do not count as tests.)
The first set of experiments involved a double control design. Bees from one hive foraged at only one station in a set of four; bees of another color foraged at all four stations. In contrast to predictions of the dance language hypothesis, recruits from both hives ended up in equal proportions at the four stations, despite the wide disparity of information presumably provided by dancing bees in the two hives. For details, see:
The second set of experiments involved only one hive. In that case, a constant set of marked foragers imbibed scented sugar solution at two stations for 3 hours each day during a 24-day period. All unmarked bees were caught and tallied during that period. On some of the days, we switched to unscented food at the target stations but provided the scented food at a third station.
On those subsequent days, recruited bees ignored the dance maneuver information in their parent colonies and arrived, instead, at the third station where we had used scented food that had been provided on the previous day. For details see:
Those sets of experiments are straightforward (with mutually exclusive designs) and the results should have been heeded. That did not happen. Instead, bee language proponents (by then, their numbers were legion) erred most seriously in two ways, errors repeated up to the present time:
1) Language advocates ignored Karl Popper’s caution: “Some genuinely testable theories, when found to be false, are still upheld by their admirers – for example, by introducing ad hoc some auxiliary assumption, or by re-interpreting the theory ad hoc in such a way that it escapes refutation.”
2) Language advocates apparently accept as valid only evidence that supports the von Frisch hypothesis. It appears that they consider counter evidence just “noise” in the system.
Back in 1865 Claude Bernard admonished against falling into such error: “[The experimenter] must never answer for [Nature] nor listen partially to her answers by taking, from the results of an experiment, only those which support or confirm his hypothesis.”
In short, during the past few decades, bee language advocates have repeatedly committed the error of only focusing on supportive evidence, while ignoring or discounting other evidence that does not fit within their “belief system.” That is, only positive evidence counts.
Flaws in “Proof” Experiments
During the past few decades we have seen the same mistakes occur repeatedly; language advocates have sought and embraced supportive evidence and dismissed negative evidence in an almost pathological manner. Any new supportive evidence always received wide acclaim, with great fanfare in the media (the exotic sells). Manuscripts submitted in support of the language hypothesis became published readily. Grant proposals that might yield positive evidence received funding. In fact, during the past several decades, millions have been spent on efforts to “prove” the language hypothesis true, once and for all.
Time in Flight and Success Rate for Searching Recruits
First came the results of some experiments by three undergraduates (Gould, Henerey, and MacLeod) that actually yielded results that did not mesh with predictions of the language hypothesis.
For example, they had marked with individual numbers a couple of thousand bees in the hive. They recorded 277 that left the hive after contacting a dancer. Of those, only 37 found either one of the two stations located in opposite directions from their hive. However, a third of those ended up at a station in the exact opposite direction from which the dance maneuver had presumably indicated the food location. Nevertheless, they concluded that attendant bees had used direction information obtained from the dancing bee. Despite that serious discrepancy, the journal (Science) would not permit us to publish a rebuttal (see Excursus EXC).
Next came James Gould’s “misdirection” experiments. By shining a bright light at an angle to dancers in the hive, he reportedly changed the direction of the dance maneuver. For his first publication in that series (Nature 1974), in support of bee language, he included the results of only three of the 33 half-hour runs he had run. While first hailed as “elegant,” it later became apparent that the small sample sizes in his experiment and repeated publication of the results of his experiments in different display formats grossly exaggerated the importance of his contribution. Neither has anyone successfully repeated his experiments.
Mechanical Bee Experiments
After people finally became disenchanted with Gould’s results, researchers in Germany and Denmark published (with much fanfare) the results they had obtained by use of a mechanical bee (“robot” bee). They also claimed that they had obtained the final resolution of the controversy. Once again, though, it became apparent that they had not resolved the matter after all. They had to use odor in their experiments in order to obtain recruitment. (Without odor, there is no recruitment, just as von Frisch observed.) In one experiment, two real foragers plying between hive and station resulted in 50 recruits showing up at the test station. By contrast, only two recruits finally arrived at the feeding station after numerous bees had attended a robot bee. Support for their conclusions gradually faded.
It is no wonder that the honey bee dance language controversy has now persisted for nearly half a century. Others conducted less exotic studies during the past few decades and claimed that they had successfully resolved the issue of bee “language.” As usual, though, they tailored their experimental designs toward gaining supportive evidence for the hypothesis – rather than conducting true tests of that hypothesis.
Radar Tracking of Recruited Honey Bees
The latest “high tech confirmation” experiments involved radar tracking of recruited honey bees. Once again we see the usual pattern: a small sample size, biased researchers, much media splash, and dogmatic conclusions. Never mind that the evidence doesn’t fit the original hypothesis; they now considered that they had obtained “proof” of bee “language.” In an interview with a science reporter for the San Francisco Chronicle, for example, the lead scientist of the radar tracking project was quoted as insisting: “‘no really sensible person can come to any other conclusion’ than that von Frisch was right.”
In their experiments the radar tracking researchers made some serious errors in approach. Consider some background. The researchers had some bees trained to fly out to a feeding station and saw them later dance. They followed a bee that attended such a dance, caught it as it left the hive, attached a transponder onto its body, released it, and then followed the path of that bee with radar. To conduct their study they embraced a number of assumptions:
1) That a bee attending a dancer has the neurological/physiological equipment to obtain abstract physical information from another bee.
2) That such a particular bee leaving a dancer “intended” to travel to the “target” site (that experimental bees had gained a route memory of the feeding station relative to the hive).
3) The capture of a bee and fastening a transponder onto it doesn’t interfere with its presumably “programmed” behavior.
4) That the released bee flight path has only one explanation (in favor of “language,” in this case).
5) That they worked in an “odor free” area (odor free to bees) for their experiments.
From the account published in Nature, it appears that some critical controls were missing, including:
1) Flight paths of bees that had been similarly treated but had not attended a waggle dance.
2) Flight paths of departing bees that had attended a waggle dance that had visited food located in a different direction.
3) Flight paths of bees searching for a scented food source (as pointed out by Rosin, above).
That is, the reasoning seems to be: if the escaping bees fly off in the “right” direction (the only direction arranged for), then they have used their language.
All of the above reminds me of a 1971 statement by Nobel Laureate Albert Szent-Gyorgyi (biochemistry): “If you know in advance what you are going to do, or even to find there, then it is not research at all: then it is only a kind of honorable occupation…”
What do beekeepers think? What percentage of them accept the assumption that one can grab a bee, attach something to its body, release it, and then expect it to go on its way as if nothing had happened? In my experience, such a bee would fly an escape path and would not likely continue a presumably “programmed” behavior. Those interested in what recruited bees normally might do after leaving the hive, including search for an odor source can access:
That publication of ours contains several quotations by von Frisch that stand in sharp contrast to the observations and claims made in this latest Nature publication.
Some other questions might have occurred by now to beekeepers and bee researchers:
1) What about the very small sample size – can the behavior of 19 disturbed bees undermine the results obtained from hundreds of unrestrained bees searching for odor sources in our double controlled and strong inference experiments? (And, note, only two bees showed up near the target station in the radar tracking experiment after some unspecified time delay.)
2) What about the behavior of the many other bees they must have experimented upon. Did they only gather data from fewer than two dozen bees and then quit when they had supportive evidence?
Tracking recruited bees with radar to study their flight paths after leaving the hive is a noble goal, provided the bees have not been disturbed. Neither do results obtained without the use of odor have relevance to the language hypothesis (as pointed out by Rosin, above). Also, the experimenters should not bother to conduct research unless they can set aside any vested interest in the outcome; that is, that they should instead execute a true test of the language hypothesis and not attempt only to seek confirmation of their prior entrenched beliefs.
As is usual in such a case, some reporters contacted me for my impressions of the Nature report. David Perlman of the San Francisco Chronicle recognized that I had not had time to study the publication and came out with a fairly accurate assessment in his article. Other reporters would not accept my explanation that I would need a few days to study the original publication, with a rather common comment: “I have a deadline to meet.”
No, I don’t think that the behavior of 19 bees will resolve this controversy, one that has run now for well more than three decades. Nor will beekeepers benefit from the claims made in this latest effort. They can go back to worrying about varroa mites, small hive beetles, and other real problems with, hopefully, some real solutions appearing in the near future. Consider the money and time spent on this supposedly final solution to the bee language controversy! Wouldn’t that time and effort have been better spent on breeding a varroa resistant bee, for example?
Despite assertions by the radar tracking crew, the exotic bee language controversy will remain with us for quite some time to come. As stated at the outset, the controversy does not revolve about evidence with respect to validity of the hypothesis. The dance language hypothesis fails to account for all the available evidence with respect to the fundamental question: “How does a naïve bee find a target food source?
Bee language advocates should now address the really important question (how do they find food) instead of attempting to prove a hypothesis. The favored hypothesis continues to be elusive. Until they address the full role of odor as first elucidated by von Frisch, we will continue at a stalemate in the controversy (see The Elusive Honey Bee Dance “Language” Hypothesis).
In summary, did radar tracking of bee flight paths resolve the bee language controversy? No, it did not; indeed, in the protocol recently reported in Nature, it could not. Because now, just as for the past thirty plus years, a statement that Patrick Wells and I published in Nature back in 1973 (pages 171-175) still holds true “. . . the honey bee forager recruitment controversy is not about the nature of evidence but rather about the nature of hypotheses. It is not what investigators observe (the data) but what they believe (infer) that is at the heart of the controversy.” See: Do Honey Bees Have a Language?
Adrian M. Wenner
967 Garcia Road
Santa Barbara CA 93103